O - Cade

12
SERVIÇO PÚBLICO FEDERAL
PROCESSO ADMINISTRATIVO N°
Z
08012.010483/2011-94
%t O .
)IGO:
80 VOLUME
REPte: E-COMMERCE MEDIA INFORMAÇÃO E
TECNOLOGIA.
REPada: GOOGLE BRASIL INTERNET LTDA.
Ç)
lGO
-------------
-02
/
/
16
03
/
/
17
04
/
/
18
05
/
/
19
06
/
/
20
07
/
/
21
08
/
/
22
09
/
/
23
10
/
/
24
11
/
/
25
12
/
/
26
13
/
/
27
14
/
/
28
DATA
/
AS MOVIMENTAÇÕES DEVERÃO SER COMUNICADAS AO PROTOCOLO
EXOS:
SEDAP/PR IMPRESSO N°47
/
TERMO DE ABERTURA DE VOLUME
Aos 29 dias do mês de Julho de 2014, procedemos à abertura do 8° volume
do Processo Administrativo n° 08012.010483/2011-94 com início às fis 1959.
Maria
~eilves Miguel
Chefe de SerR4Registro Processual
ADVOGADOS
o
.
DOCIO 04
Substitufion between offline
and online
advertising markets
Substitution between offline and online
advertising markets
Avi Goldfarb and Catherine Tucker*
November 2010
Abstract
Online advertising is increasingly subject to antitrust scrutiny, but there is a lack ofempirical
analysis on whether the large offline advertising market disciplines the online advertising
market. Here, we summarize two empirical analyses that begin to flul this void by showing an
effect of offline advertising on online advertising.
JEL Codes: M37, M38, L86
Avi Goldfarb is Associate Professor of Marketing, Rotman School of Management, University of Toronto, 105 St
George St., Toronto, ON. Tel. 416-946-8604. Email: agoIdfarbrotman.titoronto.ca. Catherine E. Tucker is
Assistant Professor ofMarketing, MIT Stoan School ofManagement, 100 Main St, E62-533, Cambridge, MA. Tel.
617-252-1499. Email: cetucker(niit.edu.
Eectronc copy available at: http://ssrn.com/abstract=1 721001
)
( fls:
L6L
1.
Introduction
Advertising markets have been turned upside down by the advent of the internet and online
advertising. The internet has grown as an advertising channel more quickly than either broadcast
or cable television, in terms of its share of existing advertising spending (Interactive Advertising
Bureau 2010). 1-lowever, what is !ess clear is how competition authorities should approach this
switch by advertisers towards online advertising. In particular, it is not clear whether or not
online advertising markets should be thought of as being distinct from offline advertising
markets.
Whether online and offline advertising markets should be treated as distinct markets is a
key question in many areas of competition authority activity. For example, the European
Comrnission, in its Google/DoubleClick and Microsoft/Yahoo! decisions, declared that, for
antitrust purposes, "online advertising is a distinct market from offline advertising" (European
Commission 2010, paragraph 61). In its statement on the Google/Doub!eClick merger, the
Federal Trade Commission makes no direct mention of offline advertising. Interestingly, the
FTC does note that "advertisers primarily purchase search advertising space to implement direct
response ad campaigns, while direct!y sold ad inventory is generaily purchased for brand
advertising campaigns" (FTC 2008, p. 7).
In this paper we discuss in non-technical terms two econometric studies which present
suggestive evidence about the extent to which offline and online markets interact. These examine
both how offline advertising influences how effective online advertising is (Goldfarb and Tucker
2010a) and how offline advertising influences the pricing of online advertising (Goldfarb and
Tucker 201 Ob).
2
Eectronic copy available at: http:llssrn.com/abstract=l 721001
2.
Identifying substitution between online and offline markets
Measuring the extent of substitution is challenging. In order to accurately measure
substitution between channels, ideally one would have a shift in price in one channel that is
unrelated to anything that might affect market conditions in the other channel. For example, an
ideal experiment would be if Google's computers, due to a mechanical error, started pricing
auctions 10 percent higher in Sweden than in Norway. Then one could potentially track how this
changed search engine advertising aliocations over time and relative to other offline advertising
media in Sweden relative to Norway.
However, generaily, there is no such ideal pricing experiment. More commoniy, we might
observe the price of offline advertising to rise in Sweden as well as an increase in both prices and
quantity soid online. It is possible this is a result of substitution between channeis: When the
price of offline advertising rises, advertisers substitute into the online channel. However, it is
also possible that this is a result of some other factor affecting ali prices: An increase in demand
for advertising might increase prices offline and online, even ifthere is no substitution between
channels. Therefore, careful analysis requires a situation in which one channel is affected but not
the other.
S
Because such situations are rare, and for a variety of other reasons, cases have iargely been
written up without reference to empirical analysis that examines substitution between online and
offline advertising channels. Ratliff and Rubinfeld (2010) discuss the dearth of empirical
evidence on how offline advertising might constrain online advertising.
Combining novel datasets with difference-in-difference econometric techniques, both of
our studies exploit variation in legal restrictions on offline marketing to infer a causal
relationship from offline marketing to online marketing. In Goldfarb and Tucker (2010a), we
3
4o
FISJ9(4L
L
examine how restrictions on offline brand (biliboard) advertising affect online brand (dispiay)
advertising. In Goidfarb and Tucker (2010b), we examine how restrictions on offline direct
marketing affect search engine advertising. The purpose of the present paper is to summarize the
resuits of those econometric studies and discuss their usefulness for guiding competition
authorities in their approach to market definition in online markets.
We recognize that the issue ofrelevant market definition is a complex and contentious area
of competition iaw and practice (Cariton 2007). For example, according to the European
Commission, a relevant product market "comprises ali those products and/or services which are
regarded as interchangeabie or substitutabie by the consumer by reason of the productsT
characteristics, their prices and their intended use."1 We also recognize that neither of our studies
meets a ciear-cut criterion such as that used in the EU of whether "customers for the product in
question can switch readily to a similar product in response to a smali but permanent price
increase (between 5% and 10%).2 In essence, our two studies exploit infinite price increases in
the offline channei and examine what happens online. The resuits are still usefui to establish
market definition, because they document that the channels are closely reiated. They do not
measure the levei of substitution as defined by the reguiation, but they shouid be viewed as
S
suggestive evidence of the existence of extensive substitution between oniine and offline
markets.
We wili discuss each of our papers in turn, foilowed by a broader discussion of the
evidence for oniine-offline substitution in other industries.
2
Official Journal C 372 of 09.12.1997
http://europa.eu/Iegislatioii sunirnaries/competition/firrns/126073 en.htm
4
3.
Goldfarb and Tucker (2010a): Billboards influence online display ad effectiveness
In Goldfarb and Tucker (2010a), we use offline marketing restrictions to identify how
offline advertising affects online advertising. 3 Specifically, we examine whether online display
advertising is more effective in locations that ban alcohol advertising on billboards.
In order to measure the effectiveness of the advertising, we use data from a !arge-scale set
of field experiments that randomized exposure to 275 different online display advertising
campaigns for alcoholic beverages between 2001 and 2008. In each field experiment, an average
o
of 223 people completed the survey, haif of whom were randomly exposed to the ad for an
alcoholic beverage and halfofwhom were exposed to an alternative placebo advertisement. We
measure ad effectiveness as the difference between these two groups in response to the question,
"How likely are you to purchase this product?" In other words, ad effectiveness is the difference
in stated intention to buy the product between people who saw an ad for the product in question
and people who (randomly) saw a different ad.
We then examine how the effectiveness of advertising differs in locations that ban
billboard ("out-of-home") advertising with locations that do not. We use two different sources of
such advertising bans. First, we compare the 17 states that ban biliboard advertising for alcohol
with the other 33 states and find that online advertising is approximately 15% more effective in
places where offline advertising is banned.
This analysis of state-level restrictions is robust to a number of controis and provides a
large-sample understanding of the differences between places with and without regulations. Still,
we cannot eliminate the possibility that there are state-level differences in advertising
responsiveness that drive the effect. Therefore, we also look at four different local regulations
This research used data supplied as part of a grant from the WPP/Goog!e Marketing Research Awards.
5
that changed over the course of our data. We find that changes in local regulations changed local
advertising effectiveness in a way consistent with the state-level effects: Ad effectiveness was
substantially higher when the ban was in effect. This result is robust to numerous controis and
econometric specifications.
These results suggest that the online advertising channel does substitute for the offline
channel. From the advertisers' point of view, they get more value from online display advertising
when the potential customers do not see any offline biliboard advertising. While we do not
o
measure the exact strength ofsubstitution between online display advertising and offline display
(biliboard) advertising, our evidence strongly suggests that advertisers should be able to
substitute relatively easily between the two channels. Still, our results do not directly touch upon
the pricing of advertising, an issue we explore in a different paper.
4. Goldfarb and Tucker (2010b): Direct marketing influences prices for search engine ads
In Goldfarb and Tucker (2010b), we examine how offline marketing affects the price of
search engine advertising.4 To do so, we exploit a quasi-experiment in regulation that restricts
the ability of personal injury lawyers to solicit clients directly—also known as "ambulance-
S
chaser" regulation. In the fifteen states with this regulation, lawyers face a waiting period before
they can contact a person after an accident. We interpret this regulation as an artificial barrier to
competition between online and offline advertising.
We combine this information on the regulation with data on the prices of location-specific
law keywords on Google in April 2007. For example, if a lawyer wanted to have an
advertisement show up prominently when a user searched for the keyword string "brain injury
This research was supported in part by a grant from the NET Institute.
attorney Baton Rouge", this would have cost $12.19 for each click received at the time we
collected our data. In contrast, the string "divorce attorney Baton Rouge" cost $7.95.
We find that the prices of search advertising for the keywords affected by this regulation
were five to seven percent higher than the prices of other keyword advertising. In particular, the
regulation affects the prices of keywords related to personal injury law but not the prices of other
law-related keywords (such as divorce law or traffic law). We show that in the states with
solicitation restrictions, personal injury keywords were more expensive relative to the price of
o
other law-related keywords in those states, compared to the price difference between personal
injury keywords and other law-related keywords in other states.
We use a number of econometric techniques to ensure the robustness of this result and we
consistently find that offline solicitation restrictions (effectively, restrictions on direct marketing)
increase the price of search engine advertising.
Prices of keyword advertisements on Google and other search engines are set using a
continuous auction. In that sense, the price changes we observe are not from deliberate pricing
decisions by Google management. Instead, the auction mechanism ensures that Google receives
something like the marginal value of receiving a click by the second highest bidder (Edelman,
Ostrovsky, and Schwarz 2007). Therefore, the observed substitution suggests either that bidders
bid more aggressively online when offiine direct marketing is limited, or that more bidders bid
online when offline direct marketing is limited, or some combination ofboth.
It is important to note that these results do not generate a cross-price elasticity estimate.
Typical market definition analysis examines the impact of a moderate price increase in one area
on prices and quantities in the other area. Our study examined an outright ban on direct mau
marketing over a period of time. Therefore, as noted by Ratliff and Rubinfeld (2010), if the
7
offline market is defined as "direct mail within 30 days" the percent change in price of the
offline channel is infinite. Ratliff and Rubinfeld (2010, p. 376) argue that therefore our analysis
is "ofdoubtful relevance to determining the boundaries of the relevant market."
While we acknowiedge that our analysis examines a ban which could be interpreted as
equivalent to an infinite price change for direct mail within 30 days, we still believe it is
informative about market: boundaries. By shutting off one (of many) types of offline marketing,
oniine advertising increased in price. The channels are cieariy related. The depth of the
o
reiationship is an open empirical question. We leave it to future work to measure whether the
exact degree of substitution warrants competition authorities to consider search engine
advertising and direct mail marketing as separate markets.
5. Online/offline substitution in other settings
Our results are consistent with an increasing body of evidence that emphasizes the
interaction between online and offline behavior in a variety of settings. Much of this literature
has focused on competition between online and offline retailers and used a similar empirical
strategy: examining how changes in offline availability affect online saies and prices. For
example, Brynjoifsson, Hu, and Rahman (2009) document substitution in women's clothing from
the offline channel to the online channel, particularly for the more popular products. Forman,
Ghose, and Goldfarb (2009) and Choi and Beli (2010) find similar effects in books and diapers
respectively. Using a slightly different empirical strategy, Prince (2007) found that an increase in
online availability of personai computer retailers led to substitution to the online channel from
the offline channel.
Because of the chalienges in finding price changes that are independent across channels,
research that measures the cross-price elasticity between online and offline markets is rare. The
8
iiterature that does exist has exploited variation in saies taxes across states to document that
peopie who live in states with higher saies taxes buy more online. in the first and most influentiai
of these papers, examining severai product categories, Gooisbee (2000) showed a substantiai
cross-price elasticity of at least 2.3, and sometimes higher than 5. More recentiy, Ellison and
Eiiison (2009), Gooisbee, Lovenheim, and Slemrod (2010), and Anderson, Fong, Simester, and
Tucker (2010) showed similar effects of saies taxes on saies ofcomputer memory, cigarettes, and
clothing respectively. Therefore, consistent with our resuits in advertising, in a iarge number of
retaii categories the iiterature suggests substitution between online and offline markets.
6. Conclusion
Whether online and offline advertising are distinct markets is an area of increasing concern
for competition authorities. In this paper, we have summarized the resuits of two empiricai
studies that document the abiiity of offline marketing to discipline online advertising. We have
aiso argued that these results are consistent with a iarge body of literature that shows substitution
between online and offline retaiier. It is chalienging to measure actual substitution between
channeis, due to the potential for spurious correiation that affects prices in both channeis
simuitaneousiy. In order to overcome these chaiienges, our studies exploit legal restrictions on
S
offline advertising. We beiieve our studies refute the hypothesis that online and offline
advertising markets operate independently and suggest a defauit position of substitution. Online
and offline advertising markets appear to be closely related. Nevertheless, we acknowiedge that
these studies do not determine the exact cross-price elasticity between online and offline ads.
They therefore shouid be treated as suggestive rather than definitive when it comes to
determining market definition.
References
Anderson, Eric T., Nathan Fong, Duncan Simester and Catherine Tucker. 2010. How Sales
Taxes Affect Customer and Firm Behavior: The Role of Search on the Internet. Journal of
Marketing Research. 47(2): 229-39
Choi, Jeonghye, and David R. Beil. 2010. Preference Minorities and the Internet. Journal of
Marketing Research. Forthcoming.
Brynjolfsson, Erik, Yu (Jeffrey) Hu, and Mohammad S. Rahman. 2009. Battle of the Retail
Channels: How Product Selection and Geography Drive Cross-Channel Competition.
ManagemeniScience 55(11): 1755-1766.
Carlton, Dennis W. 2007. Does Antitrust Need to be Modernized? Journal of Economic
Perspeclives2l(3), 155-176.
Edelman, Ben, Michael Ostrovsky, and Michael Schwarz. 2007. Internet Advertising and the
Generalized Second-Price Auction: Selling Billions of Dollars Worth of Keywords.
American Economic Review 97: 242-259.
Ellison, Glenn and Sara Fisher Ellison. 2009. Tax Sensitivity and Home State Preferences in
Internet Purchasing. American Economic Journal: Economic Policy 1(2): 53-71
European Commission. 2008. Case COMP/M.4631 Google/Doubleclick. Brusseis Mar. 12.
http://ec. europa.eu/ competitionlmergers/cases/decisions/m473 1 _200803 11 _20682_en.pdf.
European Commission. 2010. Case No COMP/M.5727 Microsoft/Yahoo! Search Business.
Brusseis
February
18.
http://ec.europa.eulcompetition/mergers/cases/decisions/
M5 72720 10021 8203 1 026 1 202_EN.pdf
Federal Trade Commission. 2008. Statement of Federal Trade Commission Concerning
Google/DoubleCI ick
FTC
File
No.
071-0170.
http://www.ftc.gov/os/caselist/
0710170/071 220statement.pdf.
Forman, Chris, Anindya Ghose, and Avi Goldfarb. 2009. Competition between Local and
Electronic Markets: How the Benefit of Buying Online Depends on Where You Live.
ManagemeniScience 55(1): 47-57.
Goldfarb, Avi and Catherine Tucker. 2010a. Advertising Bans and the Substitutability ofOnline
and Offl me Advertising. Journal o! Marketing Research, Forthcom i ng.
Goldfarb, Avi, and Catherine Tucker. 2010b. Search Engine Advertising: Substitution when
Pricing Ads to Context. Management Science, Forthcoming.
Goolsbee, Austan. 2000. In a World without Borders: The Impact of Taxes on Internet
Comrnerce. Quarterly Journal ofEconomics, 115 (2): 561-576.
Goolsbee, Austan, Michael F. Lovenheim, and Joel Slemrod. 2010. Playing with Fire: Cigarettes,
Taxes, and Competition from the Internet. American Economic Journal: Economic Policy
2(1), 131-154.
Interactive Advertising Bureau. 2010. IAB Internet Advertising Revenue Report: 2009 FuIl-Year
Results. IAB and PriceWaterhouseCoopers, April.
Prince, Jeffrey T. 2007. The beginning of online/retail competition and its origins: An
application to personal computers. International Journal of Industrial Organization 25(1):
139-156.
Ratliff, James D., and Daniel L. Rubinfeld. 2010. Online Advertising: Defining Relevant
Markets. Journal ofCompetition Law and Economics 6(3): 653-686.
0
ADVOGADOS
The road to infinity
op timiz ation
à
o
Join me for the journey,...
N PART 1,
discuss the vision behind this heady concept and why it works well to
describe the needs oftoday's marketer. 'II also play out a few examples of how
Infinito Optimization works from a consumer and marketer standpoint.
'II share how we wrapped the Kenshoo solution set around this construct as
a way to differentiate in a crowded rnarkotplace and position our clients for
ongoing success. Specifically, I'lI focus on what Infinito Optimization means in
the contoxt of Kenshoo's product offering and the problems it solves for our
clients.
o
IN PART 3,
l'II re-trace the process we went through to create the infinity loop graphic and
some of the alternate versions we carne up with along the way. You'll see how
we updated the individual components of our solution to be more in line with
our latest brand persona and idontity guidelmnes.
l'll talk about what wo're doing with the Infinito Optimization theme going
forward as it relates to Kenshoo product developrnent and marketing. We'll
explore how our lnfinite Optirnization is powered by lnfinite Innovation and
how we make the message more personal for employees, clients, partners,
and prospects.
Now, buckle up and enjoy the ride!
3
i;So
J5-
PART 1: THE Vi
liI1ihir,I.1uiiiM1uiiIir1I*isi
.
from brands to their audience.
In today's always-on world, brands and consumers are engaged in continual
twoway dialogue across myriad channels and devices. And the relationship
and advocacy goes well beyond the saIe.
As such, in the old world, media planning was easy (well, easier) as brands
simply mapped each channel to the appropriate phase in the funnel and
delivered generic messaging:
Television
i world, lanew product, and you
Radio
J
"Our product is the best, one and single... and
now you'll rernember me from thisjíngle!"
Print
"Here are directions to the location near you!"
Direct MalI
"Here's a coupon to buy this product nowl"
4
o
In today's world, it goes a little sornething more like this:
Television
"Hi world, l'm a new product, and you should be aware
of me, and find me on Facebook, and tweet about me
using this hashtag, and watch my vídeos on YouTube,
and pin me on
ri
Radio
Ii
"Our product is the best, pretty as a belie... and
now you can Google us or visit this URU'
Print
"Here's the location near you, and
a Yeip review, and a QR code, and
our Google+ page!"
Direct Mali
[crickets]
Digital
"Nice to see you, 1 know exactiy who you are,
and what you've been searching for, and what
you like, and what you're interested in, and what
content you're looking at right now; therefore 1
know this is the perfect product for you, and if
you click now but don't buy, no worries, l'll keep
following you!"
Accordingly, the job of a media planner lias never been harder. In fact, it's so
hard that it's become multiplejobs.
We have buyers, traffickers, analysts, oh my!
And that'sjust paid media. Today's marketing departments now include roles
for owned media, organic optirnization, and content marketing on top of the
traditional creative and PR roles that liave carried over from the old world.
More than anything, the difference between the old media world and the new
can he clescribed as a shift from static planning to dynamic activation.
The old world was rife with three martini lunches and insertion orders. Today's
world requires triple-shot espressos and real-time bidding.
Today's world demands careful calcuiation, constant concatenation, and
continual calibration.
Today's world requires Infiriite Optimization.
1L1II tø]iIiJti
The story of Kenshoo is essentially that of
a company that listens to its customers and
builds to their specifications.
We're fortunate to work with the most sophisticated advertisers and agencies
in the world. By tailoring technology solutions to the probiems they're facing,
we're able to create innovative products that drive tangibie resuits for ali our
clients.
Here are the rnost prevalent pain points we've heard from our clients over the
years and how Kenshoo, and specificaUy Infinite Optirnization, addresses them.
Chailenge #1
How do 1 connect with my customers?
In this world ofsmali screens and even smalier
attention-spans, marketers struggie with identifying their best custorïiers and
deiivering brand messages that truly resonate.
Solution: Closed Loop Targeting
Through Kenshoo Search and KenshoQ
S oçial, marketers can discover and
reach their best customers at the most criticai apertures - when they're
searching and when they're socializing.
Thero's a reason that search and social command 70% of online media
budgets. They're the best channeis to drive interaction and capture intent aka the 70% of the path-topurchase that matter most.
At Kenshoo, we're proud to be the undisputed leader in search and social.
Per the most recent Forrester Wave: Bid M •n.ç . ....itSo.wareProviders .04,
2012, Kenshoo was named "The Oniy Leader" in SEM technology.
And Kenshoo is the only cornpany in the world designated as a Facebook
Strategic Preferred Marketing Developer with native access to the Facehook
Exchange and Twitter Ads 1PI, so we've got serious social cred.
Throw in Kenshoo Local, the only SEM technology platform that provides
capabiiities for companies to huiid, manage, and distribute individual place
pagos across thousands of local networks, and marketers can truly dose the
ioop between online and offiine channels.
40Ei
(1
1
Challenge #2
How do 1 streamline workftow and scale my campaigns?
With more channels, pubhshers, partners, and systems than ever before, there
are too many variables in play for a more mortal marketer to manage.
Solution: Universal Integration
The Kenshoo Universal Platform providos a central place for marketers to
integrate ali their channeis and systems for automated workflow and hoiistic
campaign measurernent.
Kenshoo has completed integrations with more than 100 3rd-parties across
Search, Social, Local, Display, Affiliate, Mobile, Comparison Shopping, and
Retargeting so that we can serve as a marketer's true North for tracking and
data anaiysis.
On top of ali the various marketing channels, we also hook in with our
clients' internal systems to automate and improve campaign relevancy and
performance. This inciudes synch!:oniz...ng inventoryar other dynamic
variabies, supporting ProductListing Ads, and ty.ng. inwith cail-tra.çKipg
systems.
In terrns ofworkflow, Kenshoo empowers marketers with innovative toois to
accomplish tasks that used to take hours or days within seconds and minutes.
Exampies here include a desktop app..cation that enabiescutting and pasting
of carnpaigns across engines, func....ality to quickty find....a.çljust andsh L1tQ
any element of any campa.gn, and campaign template librariesso marketers
neverneed to startfrorn scratch.
How do our clients feei about these benefits? This Odeto Kenshoo from our
friends at Chacka Marketing about sums it up.
Where should 1 aliocate my spend?
Not ali interactions are created equal and, when managing multi-channel
campaigns, it can be difficult for marketers to know which ads are really
moving the needie against bespoke business goals.
5111il[.]iis IykiMiuTD'1T!U.i
Kenshoo SmartPath aliows marketers to map ali touchpoints along the path-toconversion and bid to the true value of each interaction.
Whether it's paid, owned, or earned media, Kenshoo can track it, determino
how much influence it had on driving conversions, and automaticaliy inform
Kenshoo's bidding systems how much each ad is actualiy worth.
By iooking at each conversion event uniquely and assigning credit to each ad
based on causality, synergy and loyalty, the aigorithms that power Kenshoo
SrnartPath deliver unprecedented accuracy.
The result is a clear view into performance by piacement and channel along
with actionable budget aUocation recommendations.
Chailenge #4
How do 1 find new ways to optimize my campaigns?
No matter how hard marketers try to stay on top of their programs, there are
always changes in the marketplace - eg, new formats, new platforms, new
competftors -- that necessitate new optimization techniques or else campaign
performance piateaus.
Solution: Infinite Optimzation
Kenshoo provides ali the essential ingredients for operaflng in an endiessiy
evolving Iandscape and effectiveiy achieving marketing objectives.
At our core are 3 components that make us better positioned than anyone
cisc in the market to deliver on the promise of Infinite Optimization - Ad.ptiv
Technoiogy, Taiiored Algorithms and UnmatchedScale.
• Adaptive Tochnology - flexible infrastructure wrapping itself around each
client to deliver peak relevancy
•
Tailored Algorithms proprietary modeis constantiy recalibrating to meet
defined go&s
Unmatched Scale - sustainable platform featuring intelligent automation to
drive maxirnum performance
Add it ali up and results are quite remarkabie.
PART 3: THE VISUAL
(You do know it and love it, right?!?)
Tine goal was to create a visual representation of the Kenshoo product
offoring. As a starting point, we blew up the old marketecture graphic we used
to describe the Kenshoo solution, which Iiad affectionately become known
interna Ily as the nuclear shelter:
q i
Â1
p''
,4L
4
'm gle
M
~
-
dck
- xO
qc 90
.
10
1
—
-
C,1
21
c,
Yande
Acm-
As you can teM, this graphic did not do our product suitejustice in torms of
displaying our offering or portrayinq the value to our clients.
So we tasked Margo Kahnrose, our inflnitely talented director of brand
management, with creating a new visual that would better tell the Kenshoo
story.
For a while, we had been toying with a planetary theme that played nicely into
our Universal Platform positioning. So our first stab at reimagining our layout
centered around orbs and rings.
Here's the fuil slideshow buildup:
The Kenshoo Story - Original Version
9
And here's where the marketecture resolved. The version below Iisted out
each ofthe search and social channels we address:
.
t.
SCARCH
:49
SUC AI
KENSØ
ZKENZHCC
1
Unkedf
4nde
Ao
KENSHDD Universal Patrnrm
o
And this version displayed the channels and systems integrated through the
Universal Platform:
KENBHØO Universal Pattcrm
crro
OOw OnMø '4N,
From there, we experimented a bit with shapes, font, and gradients to make
the key elements pop off the pago a bit more. Below is where we Ianded:
V2
KENSHDD Universal Platform
MOLL
COfMR
AO RVRS
HA
o E/4'
After shopping this around internaliy, we got some reafly good feedback from
our saies team. This depiction put too much emphasis on the platform and not
enough on the individual products.
Fixinçj this imbalance was important because we don't actualiy sell the
Kenshoo Universal Platform. Rather, the platform powers our search and social
solutions which are the systems our clients license and use to optimize their
prog ra rns.
Around this time, we also began toying with the idea of rebranding Kenshoo
Enterprise.
As Kenshoo Local expanded to inco.rporate mangenent and syndication of
page place listings through the CityGrici network, we felt it was important to
break Kenshoo Local out of our "Search Solutions" and give it an identity of its
own.
That left Kenshoo Enterprise as our sole dedicated SEM solution, so we
decided tojust cali it was it is - Kenshoo Search.
We also wanted to reframe our marketecture around the key client issues
we're solving. Our original story was too Kenshoo-centric. We needed to teu
the story of Kenshoo through our clients' eyes.
The purchase funnel is deadl Today's path-to-purchase is long and winding and
doesn't stop with the saie... it continues over the course of a lifetime customer
relationship. In fact, you might say it's infinite.
This version linked via SiideShare below was our attempt to put Kenshoo's
solution in the context oftoday's marketing landscape and this is where the
idea of the infinity loop first took hold.
Kenshoo lnfinity Loop Original Version
Oelow you can see the final frame which presents a much more simplified
version of the Kenshoo product suite, aibeit slightly incomplete. (We couldn't
figure out where to put local!)
o
íiI
We took our work-in-progress to Ybav, our CEO, and his first reaction was
that he liked the direction and dient-centric focus. He pushed us to be more
creative, though. He felt the story and graphic were too literal.
Yoav wanted to see tis visually convey the fluidity of our platform and the
magic that happens at the intersection of search and social.
So Margo and 1 went back to the drawing board... and the space therne. In our
minds, we could picture the infinity loop swirling around the search and social
planets and we were intrigued by the potential "milky way" created by the
cal isbn.
This is what we carne up with:
TheKenshooStory-SecondVeron1
We tinkered with the last frame quite a bit and eventually landed on this
version as the final resolve:
V4
KENSHGUX
KENHcDJ
Ao!
LIed
—
.-...-..--.
KENSHDD
tyirnc trbton.
This graphic definitoly felt more fluid and open. We also liked how the painterly
strokes tied into the ForresterWave, which we were still riding.
After showing it to Yoav again, we knew we were dose but not quite there. lt
was still too literal. It still felt like something that had been done before.
More than anything, Yoav felt like the shape was too solid and rigid, which
was in contrast to our agility and flexibility. It was out of sync with the continual
disruption that we pride ourselves on at Kenshoo.
After looking at the loop for another minute, ho asked, "Margo, do you paint?"
12
The final rendering of the loop closely resembies what we use today:
SOCIAL
SEARCH
LOCAL-
In later iterations, we rernoved Local (it's a different dimension than search/
social and just Iooked like an after-thought down there at the bottorn) and
hrought more color to the intersection (weaving in some Kenshoo Local and
SmartPath colors):
SOCIAL
SEARC1l
KENEHDO'
We iiked how the Rorscach-esque color spiatter in the middle emphasized that
the whoie is truly more than the sum of its parts. When search and social come
together, we can do some truiy remarkabie things such as retargeting people
on social networks with bespoke messaging based on what they've searched
for. Behoid truiy Closed-loop Targeting!
Then we created a version that includes ali 4 product logos and the channels
and systems we integrate into our p!atform to help dernonstrate Universal
Integration and Dynamic Attribution across ali touch-points.
-
4tfT*i*1
-
KENSHrch À,
KENSHOUSocial
ja 4r1'tv
i14,
KENSHEU --.
'
9
KN!HDDMAATH
So now you're in the loop with the evolution of the loop. What's next?
14
O
PART 4: THE PROMISE
It's a promise. To our dients. And to ourseivel
Ourjob is never done. The results can always improve. We can and must do
better. Ali the time. To infinity. And beyond!
Kenshoo's mission is to empower every marketer in the world with technology
to build brands and generate demand across ali media. To achieve tinis, we
need to continuaily find ways to capitalize on ernerging market trends and
create new technology solutions.
To do this, we need infinite Innovation.
IIiii1I1YLlii1
Kenshoo has a long history of first-to-rnarket innovaUon and, with our roots
in the israeii hi4och cornmunity, we have a strong heritage and culture of
innovation.
Bottom Une, innovation is notjust one of our core values... it's our core DNA.
So, how do we go about innovating?
At Kenshoo, innovation is not any one person'sjob. it's everyone'sjob.
That said, we do have one person who serves as our innovation leader aka
instigator. Her name is Danny Lev and she's shared some thoughts on the topic
of innovation on ourblog.
Danny is part of the Kenshoo Labs team which focuses on incubating new
ideas through the proof-of-concept phase before turning them over to the tech
team when they are ready to be productized.
Where do those new ideas come from?
Often times, they come from the Kenshoo
Innovation Community. This is an internai portal
i
where a wide range of ideas (679 to date!)
aro harvested from the inteliectual curiosity
of our staff and our clients Within the portal
Kenshoo ers comment and vote on oach idea
with the most releva nt ones moving forward
..,
O ..
679 ,00M Po~
As we buiid the framework for Infinite
-.
Optimization across ali media, there are some
key areas ripe for exploration and disruptíon
(aka innovation). 1 won't get into them in a public forum but, rest assured, the
Kenshoo product roadmap is aiigned with our mission and will deliver on the
true promiso of infinite Optímization.
Meanwhile, we're continuing to expand the infinito Optimization therne
throughout our marketing activity.
15
U
We first introduced Infinite Optimization at our UK Agency Summit in January.
(Of course, it was Infinito Optimisation for the British crowd.)
We got a great response from our clients and it was fun to soe them talk about
what Infinite Optimisation means to them:
jKenshooDeUverslnflrute Optimisation
l
In March, we updated Kenshoo.çom to put Infinito Optirnization front and
center with the Une, "We deliver unparalleled results by empowering
sophisticated marketers with cutting-edge technology."
We also began to expand on the infinity concept and loop shape, thinking of it
more as a palette upon which we can paint a portrait of our promise.
Here's some pretty prose to that effect wrítten by Margo:
"Superior research and engineering combined with a creative/innovative
attitude puts Kenshoo squarely at the intersection of art and science and
at the helrn of each. Kenshoo is future-focused; our clients are exploring new
frontiers with infinite momentum. The possibilities are endless.
The infinity loop is reprosented artistically to demonstrate the harrnony and
synchronicity between various aspocts of our offerings, and the 'artisans'
behind thern.
The graphic can house correlative pairs and symbolize the fluiclity of motion
and continual improvement between them, i.e. Marketers & Technology, Goals
& Results, Brands & Consumers, Local & Global, Intent & Interaction, Search &
Social, etc."
16
The next step was to make the promiso of Infinite Optimization moro personal.
To begin, we equipped everyone at Kenshoo with email signatures and
business cards that included a fihl-in-the-blank to show what motivates Lis.
Aron Gidmrn
Mtrkin Offcxtr
O1ce -1877 56-7482 1O
CIO
KENSHDO
infinite Optimzttlin pcwrn7d by in1nibt tnotgy
KeHy Wrather
Munagor, Conte MarkeWig
cc, 1
Ofike 1 77
C /4C1 1/2
KENSHDQ
o
POWERED BY NFIMT
1bJ N M
shoo.norn
Avne, Ste 400 Chnoc,
M í5 M
ÍÂ KENSHOD
We also had a video contest to soe what fueis Infinito Optimization for each of us:
Infinito Optimization Video Contest
1
We took the brush strokes that represent the infinity loop and swathed thom
across our coateral and social media profiles.
We also added painterly touches to our print adverts:
't:e/ ha. ,, Ia/e to lei!.
1
1
1'C( 11
i//LífL?
i 'C1A.
YOUR A0)IENCE.
— -KENSHDO
Por
oo .. Yetlertnç nec
17
oo st,flon
)
More recenfly, we've installed scribbe walis in offlces for people to share daily
doses of nspiraton:
And we had famous lsraeli artist, Ram. Me1h,
hod workshops with ourteam to
1
create works of art rnodoled off our core values.
.1
18
ou
The latest Infinite Opumization offshoot is a media campaign we're running in
digital and print. (Other channels and formats coming soon!)
Here we're applying the art and science theme a bit more litorally...
Digital marketing is an art and a scionce.
The key to success is striking the right balance.
Our clients are talented artisans that thrive on insight and clarity.
Kenshoo'ers are crafty scientists building the elite tools of the trade.
The result is that each artist can unleash his/her talent to achieve fuli potentiai
and create ROI masterpieces.
Here's an ad we ran in the Digital MarketingDepot Guide to PPCCamp
Managernent Tools:
r
Digital rnarketing is an
An artist's work is never finished.
The right toots make the masterpioce.
KENSHOD
Poweririçj Mzstnrpipcç.,s for
Q TRE FORTUNE 50 and ALL iO TOP GLOBAL AD AGENCY NETWORK$
And here's something we ran onnsideFacebook.corn and AllFacebook.com.
Th Key to Successjül
Social Advertisfng
CREATE AN ROI MASTERPIEcE
19
Live
Well, folks, it's been an incrediblejourney...
Forget the 8 months we've spent at Kenshoo bringing this idea to life, can you
believe we made it through more than 4,500 words here?!?
haven't written this much in one place since my book. Well, maybe an RFP
response:)
So where do we go from here?
From a product standpoint, we're committed tonfiçiie Innovation
.. to offer
opportunities for our committed clients to perpetually perform.
From a marketing standpoint, we plan to infuse more art mio the mix and find
new engaging ways to hetp our clients and the community at large wrap their
arms around this concept and make it their own.
lndeed, at the K8 Summit, Kenshoo clients and partners from around the
world gathered in Sausalito, California to participate in a day of interaction and
innovation. For a recap ofthe event, you can read a post on theKenshooblog
by Lmndsay Kleinick, our events manager.
One ofthe highlights ofthe day was the Kenshoo Ecosystern Challenge. This
two-hour activity took place during the afternoon and gave attendees a chance
to explore the beautiful resort grounds at Cavallo Point while engaging with
various Kenshoo partners to learn how they support holistic digital marketing
prog ra ms.
To help teams navigate the scavenger hunt, we armed each team with an
iPad mini and a cool app designed by Mint Chip. The app guided participants
through 16 checkpoints and provided information about each Kenshoo partner
and the value they add to our ecosystem. (Yes, there was a quiz at the end!)
o
One ofthe stations was led by farnous lsraeli Artist, R
.P.Mi Meiri. Rami asked our
clients to dig deep and add to the canvas a portrayal oftheir inspiration. From
travei to nature, we got some great contributions and the final result was truly a
masterpiece.
Lfl
And at cimexco in Cologne, we engaged clients and partners with our Infinite
Optimisation stand:
We've got more in storo as we carry (and tweak) our message across the globe
to a location near year! 1 hope you enjoyed this inside look at Infinito Optimization
and how it carne to be.
And 1 hope you'll hold us to our promise of delivering Infinite Optimization at
every turn along the road...
n
21
POST SCRIPT:
THE DEFINITION
This originaily ran as a series on the Kenshoo jçg: in the Loop.
Over the span of 16 posts, 1 covered infinite optimization theme from every
angie and (more appropriateiy given the infinity ioop shape ... ) curve.
.
We broke down the executive vision, market chauenges, design progression,
ad campaigns, and future promise.
Somehow, through it ali, 1 never gave a singie, succinct sentence deflning
Infinite Optirnization. Hat tip to my man, Jsh, for pointing this out.
Landing on the definition was itseif a process of Infinite Optimization. #someta
Here are some of the versions 1 cycled through...
Infinite Optirnization is the solution Kenshoo delivers for building brands and
generating dernand through digital marketing. (Connects with ouLí. ission
statement.)
Infinite Optimizatiori is Kenshoo's unique approach to helping advertisers and
agencies improve digital marketing performance. ( Expiains who our clients are
and what we help thern do at a higher levei.)
Infinita Optimization is Kenshoo's unique approach to marketing in the rnodern
era. (A bit more provocative.)
•
Infinito Optimization is Kenshoo's solution for helping marketers mLlst navigate
a complicated web to acquire, retain, and grow custorners. (A bit more
sol utions-focused.)
Here's where 1 landed...
Infinite Optimízation is Kenshoo 's solution for continually irnproving marketing
performance. (K.l.S,S.)
1 floated the idea out to a few team members who suggested combining a few
to realiy drive the point home.
So 1 carne up with this...
Infinite Optiinization is Kenshoo's unique approach to continually improving
performance in a complex marketing landscape.
Then Keliy offered up this verbiage she wrote for an upcoming S predfast Ljog.
post...
"From the Kenshoo perspective, this idea of optimizing the entire customer
journey and continuaiiy refining the process to achieve optimal resuits is what
we cali Infinita Optimization - a true ciosedIoop experience."
22
Iiked liow this version tied in the customer viewpoint and emphasized results.
So 1 formed ths definition...
Infinite Optimization is Kenshoo's solution for nnarketers to continually improve
evory step ofthe custornerjourney and cichieve optirnal results.
Then Josh woighed in that Infinite Optimization is reaHy more of an ideal than a
solution or approach. He suggested framing it as a shared goal, nay, the shared
goal.
So here's what we're going with (for now)...
Infinite Optimization is the shared goal between Kenshoo and
our clients to continually improve every step ofthe customer
journey and drive optimal marketing performance.
And the road goes on for, ver...
23
ADVOGADOS
o
Onfine Advertising: Defining
Relevant Markets
Market Definition in Online Advertising
1. INTRODUCTEON
The rapid growth of the Internet, and the incredible flow ofinformation that the Internet has made possible,
has lransformed the business of advertising.' Today it is difficult to surf the web without seeing online
advertising, ofien in the form of visual display ads on web sites (including pop-ups and pop-downs) and
textual ads on search sites.2 There is little doubt that on-line advertising has taken business away from
traditional modes of advertising, such as newspapers, snail mail, and radio. What is !ess clear is whether the
shift is price driven and whether traditional advertising channels constrain the pricing of Internet ads.
This paper provides an overview of lhe development of Internet advertising. In the process, we describe
lhe nature of advertising competition as it currently exists onhine. We focus on lhe extenl to which various
types ofonline advertising compete with each other and with offline advertising. While our goal is not to
reach a definitive opinion as to how relevant markets ought to be defined, we do suggest a number of core
empirical questions whose answers will help to clarify questions surrounding market definition.3
.
The paper proceeds as follows. In Section II, we describe lhe birth and growth of the Internet and onhine
advertising. Section III offers a broad overview of both onhine and offline advertising and lhe economic
models that allow one to evaluate competition among advertisers. In Section IV, we focus on online
advertising and distinguish lhe various types ofonhine ads and lhe means by which those ads are marketed.
Section V focuses on competitive issues. We evaluate the extent to which onhine and offline ads compete
and we also ask whether various types of onhine ads are competitive with each other. In Section VI, we
offer some brief concluding comments and suggestions for further research.
II. TUE ORIGIN AND GROWTH OF ONLINE ADVERTISING
A. The Birth and Commercialization of the Internet
The inteliectual and technical underpinnings of lhe Internet go at Ieast as far back as the very early 1960s,
when MIT's J.C.R. Licklider coauthored a trilogy of memos describing lhe "Galactic Network" concept4
and Leonard Kleinrock, also of MIT, published the seminal paper on packet-switching theory.5 These
cornerstones led in 1972 to the first public demonstration of the ARPANET—the precursor of today's
Internet—and tolhe introduction ofelectronic mau. By 1985, the ...
Internet was already well established as a technology supporting a broad community of
researchers and developers, and was beginning to be used by other communities for
daily computer communications. Electronic mail was being used broadly across several
communities, often with different systems, but interconnection between different mail
systems was demonstrating the utility of broad based electronic communications
between people.
Lower-case-i "internei" originally referred to any nelwork of networks. (Debra Littlejohn Shinder, Computer Networking
Essentiais, Cisco PRESS, 37 (2001).) Upper-case-I "Internet" refers to "the global informatiori system that is logically linked
together by a globally unique address space based on the Internet Protocol (IP)...... (Federal Networking Council, FNC
Resointion: Definition of "Internet, "(October 24, 995), a! http://www.nitrd.gov/fnc/lnternet_res.html). Usage has been moving
in the direction of using lower-case-i internei to refer to the global network. (See examples at internei capilalization conventions,
at http://en.wikipedia.org/wiki/lnternet capitalization conventions).
The beneflis of advertising can also be achieved when information about the business appears on the list of "organic resulta"
displayed by lhe search engine.
Market definition is a means to an end—to a competitive analysis of a merger or of a non-merger activity. As a result, a market
definition exercise outside lhe merger conlext will sornetirnes deviate substantially from the exercise that would be undertaken if
there were a merger.
J.C.R. Licklider, Man-Computer S rnbiosis, (HFE-I) IRE TRANSACTIONS ON HUMAN FACTORS IN ELECTRONICS 4-11 (March
1960) ai http:/Igroups.csail.niit.edu/medglpeople/psz/Licklider,html, J.C.R. Licklider and Welden E. Clark, On-line nian-computer
conunun/cation, Proceedings of lhe May 1-3, 1962, Spring Joint Computer Conference, AFIPS JOINT COMPUTER CONFERENCIES,
a! hltpífportal.acrn.orgicitation.cfrn?id=1460847; J.C.R. Licklider and Robert W. Taylor, The Computer as a Comniunication
Device, SCIENcE AND TECHNOLOGY 21-41 (April 1968), ai http://www utexas.edu/ogs/lectures/taylor/licklider-taylor.pdf.
Leonard Kleinrock, Information Fiow in Large Co,n,nunication Nets, RLE QUARTERLY PROGRE5S REPORT (July 1961).
Barry M. Leiner, Vinton O. Cerf, David D. Clark, Robert E. Kahn, Leonard Kleinrock, Daniel C. Lynch, Jon Postei, Lariy O.
Roberts, Stephen Wolff, A Brief History of lhe Internet (3.32), (December 10, 2003) [hereafter "Leiner et ai. (2003)"], ai
http://www.isoc.org/internet/history/brief.shtinl.
:ti D
(FsJ3,r
Market Definilion in Online Advertising
".------,
At this point in time the Internet was literaily not open for business. The National Science Foundation
(NSF) operated the Internet's nationai-scale "Backbone" and enforced an "Acceptable Use PoIicy" (AUP)
which prohibited usage for purposes "not in support of Research and Education."7 Not until 1993, when the
NSF reinterpreted the AUP, was the Internet fuliy opened to commerciai traffic.8
II. The Importance ofthe World Wide Web, the GUI and the Browser
The eariy-1990s Internet provided connections between sites, and individuais at various sites had
information and resources that would be useful to others. But discovering and sharing that information was
a daunting chalienge. There was no easy or systematic way to uncover what information was available
where or how to access it. Coliections of information were balkanized, uncataioged and unindexed, and
cloaked behind cryptic file narres. Users maintained personal lists of what they had found, or learned of
through word ofmouth, in their imperfect views into the Internet.9
.
New information management systems such as Gopher and Wide Area information Servers (WAIS)
were created and were signiticant improvements—but feli short of what was needed.10 It was Tim BernersLee's conception and development of the World Wide Web as a decentralized, scaiabIe system ofhypertext
links that calaiyzed the revolution that the Internet has become.1 '
Marc Andreessen and Eric Bina from the National Center for Supercomputing Applications (NCSA) at
the University of Illinois released the Mosaic web browser in 1993—the flrst browser that ailowed for the
dispiay of photographs and graphics positioned within a page of text. Andreessen cofounded Netscape in
mid- 1994, releasing what became the Mosaic Netscape (later, Netscape Navigator) browser for ali major
piatforms on October 13, 1994.12 Millions of users took advantage ofNetscape's browser, which quickiy
grew to be the most popular browser in the market.
C. Directories and Search Engines Increased the Value of the Web
The deveiopment of the web and of browsers did not by itseif solve an oider probiem: consumers couId
become aware of other sites on the Web oniy by word of mouth (e.g., sharing "hot iists") or through
reconimendations from other sites (e.g., Cooi Site of the Day). WebCrawier, iaunched in 1994, was perhaps
the first search-engine service that embodied the three fundamentais we now expect: it was crawier based
(to discover new sites), indexed, and abie to search the fuii text (not just tities or summaries) of sites.13
There rapidiy foiiowed a proiiferation of search engines, such as Lycos, Mageilan, Excite, lnfoseek,
inktomi, and AltaVista. Yahoo! took a different approach, using its "staff of experts" to categorize web
sites into a hierarchicai structure to buiid a directory around subject-based, demographic, and geographic
content.
In 1996, Stanford graduate students Larry Page and Sergey Brin began a research project that uitimately
became a patented innovation in search and the beginnings of Googie.'4 Googie remains lhe most-popular
search site today. In September 2009, Americans conducted almost 14 muiiions searches. Aimost 65% of
lhese searches were conducted on Googie's sites. Yahoo had the second most popular search engine,
Leiner eI ai. (2003).
Worldwide Web Consortiuin (W3C), A Little Histo,y o! lhe World Wide Web, ai http://www.w3.org/History.htmi.
I.R. 0km, THE INFORMATION REVOLUTION THE NOT-FOR-DUMMIES CUIDE TO THE HISTORY, TECHNOLOGY, ANO USE OF THE
INTERNET (lronbound Press, 2005).
Neither Gopher nor WAIS used hypertext. WAIS connected only search engines togelher. Gopher's prospecls were darnaged
when lhe University of Minnesota announced it would charge a license fee for Gopher to certain classes of users. [Tini BemersLee and Mark Fischetti, WEAvING THE WEB 72-74 (HarperCoilins 1999) (hereafter "Berners-Lee (1999)"].)
For the story of lhe developrnent of the World Wide Web, see Berners-Lee (1999). To avoid lhe mistake made with Gopher,
CERN pledged that lhe Web prolocol and code would be available free of charge to ali users and uses.
The Netscape browser was "free but not free." II was free for sludents and educators and theoretically $39 for alI others, though
this was not enforced. Internet Pioneers: Marc Andreesen, ai http://www.ibiblio.org/pioneers/andreesen.html.
" WebCraivler 7'irneline at http://www.thinkpink.comlbp/WebCrawler(History,hlmi.
' .iohn Battelle. THE SEARCH: How 000GLE AND ITS RIvAL5 REWROTE THE RULES OF BUSINESS ANO TRANSFORMED OUR
CULTURE Chapter 4 (Penguin Books 2005).
8
Markei Defin/ilon in Online Advertising
processing almost 19 percent of ali searches. Microsoft's new Bing search engine claimed over 9 percent
ofsearches, and its share has been increasing since its launch in May 2009.16
D. The History ofAdvertising on the Internet
There appears to be no consensus on preciseiy when advertising on the Internet began. Tim O'Reilly,
founder of the web portal Global Network Navigator (GNN), claims that the first advertising appeared in
1993 on GNN and required "special dispensation from the National Science Foundation."17 Others cite a
banner ad sold to AT&T and displayed on the HotWired site in 1994)8
At first, online ads were sold exclusively on a cost-per-impression ("CPM") pricing model used by
offline media, i.e., the advertiser was charged proportionally to the number of times the ad was displayed
on a web page. That changed in 1998, when the search engine GoTo.com was launched.19 GoTo.com broke
with cost-per-impression pricing, instead auctioning the top resuits of its search-result pages, with
advertisers' sites appearing in descending order of their bids (on a pay-per-click basis).20 GoTo used a realtime competitive-bidding process to aliocate listing priorities. More specifically, GoTo's process was a
"first-price auction" in that the winning bidder paid the amount of its bid for every click.2' Edelman ei ai
(2007) provide an illustration of how GoTo's auction mechanism was "far from perfect." They note that
GoTo and its advertisers "quickly iearned that the mechanism was unstable due to the fact that bids could
be changed very frequentiy."22
Google launched its AdWords service in October 2000; the service piaced ads on the search-results
pages on googie.com. The ads displayed were chosen based on the keywords that appeared in the user's
search inquiry. Google aimed for low transaction costs: AdWords was described as "self-service," allowing
sign-up, activation with a credit card, and ad design and iniplementation from the Google web site. These
text ads were sold on a cost-per-impression basis, for 1.5Ø, 1.20, and 1.00 per impression "for the top,
middle, and bottom ad unit positions, respectively."23
Google updated its AdWords program in February 2002, introducing AdWords Select, a program that
used cost-per-click pricing.24 An advertisement's ranking was based on a combination of the advertisers
°
.
7
20
°
22
24
cornScore, comScore Releases September 2009 U.S. Search Engine Rankings, (October 14, 2009), ai
http://www.eoniscore.com/Press_Events/Press_Releases/2009/1 0/comScore_Releases_September_2009_lJ. S._Search_EngineRa
nkings.
eMarketer, Anaivzing lhe Bing Effect, (September 29, 2009), ai http://www.ernarketer.corn/Article.aspx?R=l 007297.
"The first internet advertising appeared in 1993, foI 1996. 1 know, because 1 did it, under special dispensation from the National
Science Foundation, on our pioneering web portal, GNN, or the Global Network Navigator. GNN was sold to AOL in 1995 and
soon withered away there, but it was the first comniercial, ad-supported web site, and launched the first web ads in late 1993."
(Tini O'Reilly, cornments on Paul Kedrosky, Updaled: The Firsi Decode of Internei Advertising, (March 7, 2007), ai
http://paul.kedrosky.comlarchivea/2007/03/07/the first _decad.htrnl.)
See
also
O'Reilly
biography,
ai
http://oreílly.com /oreilly/tim bio.csp: "1993. O' Reilly's Global Network Navigalor site (GNN, which was sold to America Online
in September 1995) was lhe first web portal and the first true conimercial site on the World Wide Web." Wikipedia more
specifically claims that lhe first Internet banner ad was sold by GNN to HeIler Ehrman LLP. Global Network Navigatar, ai
http:/Ien.wikipedia.org/wiki/GlobalNetworkNavigator.
Barbara K. Kaye and Norman Medoff Jusi a Click Away: Advertising on lhe Internei, MASSACI-IUSETTS: ALLYN AND BACON
(2004). (Cited by David S. Evans, The Online Advertising Indusiry: Economics, Evolution, and Privacy, (23:3) JOURNAL OF
ECONOMIC PERSPECTIVES, 38 (Surnmer 2009).)
GoTo renarned itselfto Overture Services in 2001 and was acquired by Yahoo in 2003. Danny Sullivan, GoTo Makes Overture To
New Nome, SEARCI-I ENGINE WATCH (October 2, 2001), ai http://searchenginewatch.corn/2 164231; Yahoo, Yahoo! To Acquire
Overture, (July 14, 2003), ai http://docs.yahoo.corn/docs/pr/rel case] 1 02.html.
,leff Pelline, Pav-for-placement gela another shot, CNET NEWS (February 19, 1998), ai http://news.cnet.com/Pay-for-placernentgets-another-shot/2100-1023_3-208309.html; Danny Sullivan, GoTo Going Sirong, SEARCH ENGENE WATCH (July 1, 1998), ai
http//searchenginewatch.com/216633 1.
An alternative is a "second-price auction"--related to the auctions run by Google and Yahoo today—in which the highest bidder
wins lhe auction but, ínstead aí paying its own bid, lhe winner pays the secondhighest bid.
Benjaniin Edelman, Michael Ostrovsky, and Michael Schwarz, Internei Advertising and lhe Generalized Second-Price Auction:
Se/ling Billions ofDollars Worih of Keywords, (97:1) AMERICAN ECONOMIC REvIEw 246 (March 2007).
Google, Google Launches Self-Service Advertising Program (October 23, 2000), ai
http://www.google.com/press/pressrel/pressrelease39.htrnI. See also Google, Google Milestones, ai
http://www.google.coin/intl/en/corporate/history.htmi.
Google's "Prernium ads," that appear on lhe top of the search-results listing continued to be sold on a cost-per-impression basis
for severa] months. Subsequently, lhe AdWords Premium and Select prograrns were nierged.
Markei Definition in Online Advertising
per-click bid as well as lhe ad's click-through rate.25 By March 2003, Google had over 100,000 advertisers
buying search ads through its AdWords program. Google was then serving 200 miliions searches per day
on ali of its sites worldwide.26
In March 2003, Google broadened away from the search-resuits pages and began offering ads "to the
rest of the web." Ultímately called "AdSense," Google's new program was contextually targeted (Le., it
matched advertiser keywords to the "meaning of [the] web page" on which the ads would be displayed) and
used lhe click-through rates of ads as a determinant of the prominence ofplacement that would be given to
the ads. According to Google, "[u]sers see the most relevant advertising first and advertisers are rewarded
with average click-through rates at least five times higher than the industry average for traditional banner
„
ads.
Google introduced "Site Targeting” in April 2005. The software was launched as a beta feature that
enabled advertisers to aim their ads directly at particular web sites in the Google Network. Advertisers
using these site-targeted ads can use animated images (that had not previously been allowed) in their
advertisements in addition to text and static-image ad formats.28
.
in March 2009, Google began a beta test of a new ad-targeting system for its non-search, AdSense ads.
Until that time, Google had chosen the ads it displayed on a web site owned by an AdSense partner on the
basis of a match between the advertiser's selected keywords and the content of the web sites on which the
ads were to appear. However, under this new, "interest-based advertising" method, Google takes into
account additional information about lhe user's browsing history. According to Google, "[t]hese ads will
associate categories of interest—say sports, gardening, cars, pets—with your browser, based on lhe types of
sites you visit and the pages you view. We may then use those interest categories to show you more
,29
relevant text and display ads.
Google introduced AdSense for Mobile in September 2007, which aliows owners of web sites optimized
for mobile devices to monetize those sites by allowing Google to display AdSense text ads on them.3°
Moving further in the mobile direction—and away from the web—in June 2009 Google released a beta
version of AdSense for Mobile Applications that pays developers when ads are shown in iPhone and
Android appiications.3 '
III. THE ECONOMICS OF ADVERTISING
A. The Objectives of Advertising
Companies typicaily advertise to achieve one or more of several possible goals: to inform, persuade, or
remind, or to build brand awareness or brand loyalty. Successful advertising can !ead to increased saies
and/or a reduction in the price elasticity of consumers' deniands for the advertised product. Either can
increase revenues and profits (ifthe incremental profits outweigh the incremental costs of adverti sing).32
lfads were prioritized solely on the basis oftheir bids, an advertiser could bid a high per-click amount on an ad that users were
uninterested in clicking. This would resuil in low revenue for Google. Moreover, it would also be likely that users did not find the
low cl ick-through-rate ad relevant. By including click-through rate as a deterininant of lhe ad's positioning, Google increases its
revenue, while helping its users to see ads with high relevancy.
' Google, Google Builds World's Largesi Advertising and Search Monet izalion Program, (March 4, 2003), ai
http://www.google.corn/press/pressrel/advertising.htrnl.
27 Google,
Google Builds World's Largest Advertising and Search Monetization Progran,, (March 4, 2003), ai
http://www.google.com/presslpressrel/advertising.html.
28 Google, Sue Targeting, (April 25, 2005), aí http://www.google.corn/press/annc/sitetargeting.html.
2" Susan Wojcicki, Making ads more interesting. THE OFFICIAL 000GLE BLOG, (March II, 2009), ai
http'/lgoogleblog.blogspot.cornl2009/03/rnaking-ads-more-interesting.html.
Alex Kenin, Here comes mobile, INSIDE ADSENSE, (September 17, 2007), ai http:l/adsense.blogspot.com/2007/09/here-comes111obile.html.
n Susan Wojcicki, Announcing the AdSense for Mobile Applicaiions beta, THE OFFICIAL 000GLE BL0G, (June 24, 2009), ai
http://googleblog. blogspot.com/2009l0ôfannouncing-adsense-for-rnobile.htrnl.
32
See Robert Pindyck and Daniel Rubinfeld, MTCROECONOMICS § 11.6 (716 ed. Pearson 2009) for an introductory discussion of lhe
strategies involved in niaking advertising decisions.
Market Definition in Online Advertising
The saie of advertising to businesses and the dispiay of advertisements to consumers take piace in a twosidedmarket at the hub of which sits the content publisher (and any other intermediaries facilitating the saie
and/or dispiay of the advertising). 3 '34 The publisher's function is to rnatch consumer eyebaiis with the
rnarketing messages of businesses; the publisher profits when it is able to attract the consumer eyebaiis at a
cost iess than the amount the businesses are willing to pay the publisher to dispiay their ads to these
consumers.
In many two-sided advertising scenarios, the profit-making side (the advertisers) must subsidize the
consumer side, where the subsidy to the consumer is the provision of the non-advertising contenttypicaily for free or at ieast beiow the average cost of producing and distributing the content—that attracts
the consumers in the first piace. The publisher pays for the creation and distribution of the content from the
revenue it receives from the advertisers.35
B. Targeting ofAdvertisements
Effective and efficient targeting of ads, whether on- or off-fine is irnportant not oniy to advertisers, but also
to those who have advertising opportunities to seu. In a world in which advertising spots are being
elTectively placed and soid, advertisers and those that piace the ads wiii each profit from the effort. To
determine the extent to which advertising is iikeiy to be profitable, advertising must take into account the
fact that consumers wili vary in their receptiveness to any given advertiser's message. For a given number
of ad exposures, therefore, the advertiser can expect to have better outcomes (e.g., saies) the more closeiy
those exposures are targeted to consumers that are receptive to the advertiser's message.
As a general ruie, the cost of an advertisement can be expected to be a function of the exposure that the
advertisement is iikely to get. For offline ads, such as newspaper ads, the price to the advertiser is typically
a function of the number of times the ad wili run and the number ofpotentiai customers that are likeiy to be
exposed to the ad. With online advertising, the sarne general principie appiies, but the firm placing the ad
has greater pricing fiexibility, being able to charge for example according to the number of customers that
either click on an ad or are otherwise exposed to the web page with lhe advertisement for a significant
period of time.
In either instance, if the advertiser pays for exposures, for a given price per exposure the advertiser
prefers targeting those exposures to consumers that are more iikeiy to becorne customers. The result is that
the better targeted the exposures, the greater the wiiiingness to pay of the advertiser. However, for a given
price per exposure, the selier of advertising is in the short run indifferent to the advertiser's results (because
the seiler gets paid for exposure not for results).
.
Because advertisers have higher wiiiingness to pay for better-targeted ads, in the long run lhe seiier of
the ads wiii prefer being able to better target the ads it sells in order to increase the value to advertisers of
the selier's inventory of advertising opportunities. With offline advertising, targeting rnight invoive
geographic scope in the case ofnewspapers, or time ofday in the case of radio advertising.
With either offline or online advertising, ifthe advertiser pays for results (e.g., on a per-click basis for
oniine ads), it is the se/ler of advertising that prefers that the ads be weii targeted to iikely customers.36
With oniine advertising, when paying per click, an advertiser rnay be indifferent between (a) a smaii
number of exposures to a weil-targeted group of consumers (i.e., yielding a relativeiy high click-through
" Sinion P. Anderson and Jean J. Gabszewicz, The media and adveriising: a fale oftwo-sided ,narkets, HANOBOOK OF CULTURAL
EC0N0MICS (Victor Ginsburgh and David Throsby, eds. forthcoming).
34
Other examples of two-sided rnarkets include credit-card nelworks (matching consursiers bearing the network's cards with
rnerchants that accepl lhe network's card) and shopping maus (that match consumers wishing to niake a variety of purchases with
a variety of merchanls wishing to sell their wares). Jean-Charles Rochet and Jean Tirole, PIaiforn1 Competition in No-Sided
- Markets, (1:4) JOURNAL OF THE EUROPEAN Ec0N0MIc ASSOCIATION 990-1029 (June 2003).
" The subsidy is necessary in such settings because the advertisements are often insufficient draws for consumers' eyeballs (or lhe
ads are seen by lhe consumers as a negalive that must be endured as part of the deal to receive desirable content for free) Of
course, lhis is nol always lhe case. For example, with "want ads" consumers seek out advertisenients precisely in order to oblain
lhe included information. !ndeed, lhere are instances in which advertisers are willing to pay to have their ads displayed and
consumers would be willing to pay to see those ads.
Here clicks are an iniperfecl proxy for purchases, which represent lhe ultimate objective of the advertiser. We are implicitly
assunling that purchases derived from the ad are proportional to the number of clicks on lhe ad.
Market Dej7nition in Online Advertising
rate) or (b) a large number of exposures to a less-well targeted group of consumers (i.e., with a relatively
low click-through rate).37 in this way, the advertiser that places an online advertisement is insured against
the possibility of a relatively unresponsive group of individuais who receive exposure to the ad.
The selier of advertising, however, faces an opportunity cost for every impression sold to an advertiser.
If ads sold to an advertiser deliver a relatively low rate of click-throughs, the seiler would have preferred
either (a) to have sold those impressions to an advertiser whose message would have produced a higher
click-through rate or (b) to have deiivered the advertisement to a different group of consumers—consumers;
that would have generated more click-throughs.
To sum up, whether the advertiser pays for exposure or for results, the selier is better off if it can deliver
its ads to a welI-targeted group of consumers. Weli-targeted ads make advertisers better off if the
advertisers pay for exposure; pay-per-click advertisers are at least as weli off when ads are better targeted.38
.
Advertisers have the capacity to target consumers on many dimensions. A particular advertiser might be
interested primariiy in a demographic cross section of consumers, perhaps defined by a range of income,
age, gender, or ali three. A diff'erent advertiser might be strongiy interested in consumers with a particular
interest or hobby—with only a second-order interest in demographics. A third advertiser might find neither
demographic nor interest/hobby targeting effective but would instead want to target consumers with a
demonstrated current need for the advertiser's product or service (i.e., the consumer is actively searching
right now for that advertiser's product).
Advertisers whose products or services are of broad appeal can be expected to be less interested in
targeting the group of consumers to whom they advertise. Advertisers whose products or services are of
narrow appeal are more likely to value relatively highly the ability to target specific groups of consumers.
C. Choosing Media Outiets
For a given value of a saie, an advertiser whose product is ofbroad appeal will likely eschew media that
could be highly targeted because such media would be more valuable to advertisers that value that targeting
ability (and therefore the broad-appeal advertiser might essentialiy be outbid by the narrow-appeal
advertiser for highly targeted advertising).
S
There are many different media available for advertising, and many particular vehicles within each class
of advertising media.39 In addition to deciding how much to spend on advertising in general, an advertiser
must also decide how to allocate its expenditures among those media and among their vehicles.4° The
aliocation decision is a complex one because advertising media vary along several dimensions that reflect
differential abilities of ads to achieve the advertisers' desired objectives. For example, a giossy magazines
ability to reproduce high-quality color images, coupled with the "coffee table effect," might make it very
weli suited to brand buiiding; whercas the magazine's long lead time, and its non-urgent consumption by
their readers, could make this medium poorly suited for an advertiser needing a customer response in a
" There niay be reasons, likely second order in importance, as to why an advertiser paying on a per-click basis might slrictly prefer
its ads be shown to a better-targeted audience. For a gíven textual message in the ad, a click by a less specificaliy targeted
consumer might be a weaker signal of the consumer's propensity to buy the advertised product than would a click by a more
targeted consumer; thus a click by a better-targeted consumer would be more valuable. Some consumers nay click on an ad Out of
curiosity but with no interest in purchasing. it is possible that a better-targeted group of consumers would produce a lower
incidence of curiosily-driven clicking.
An advertiser may need to broaden its nel beyond highly targeted consumers in order to obtain more responses to its ad. For
example, a retailer selling a seat back organizar for use on airplane flights bought search ads triggered by the queiy "airline seal
back organizer" for a nickel per click. However, lhe retailer found that these ads produced few clicks because foi enough
consumers knew this product category existed. The retailer considered advertising in response to lhe more-popular search phrase
"travei accessories" but found that lhe higher per-click price to do so, $1.50 at that time, was prohibitive. (Darren Dahl, Real-Lfe
Lesson.s
in
1LJS/ng
Google
AdWords,
NEW
TIMES
Y0RK
(October
14,
2009),
htlp://www.nytimes.com!2009/l 0/1 5/business/smallbusiness/l 5adwords.html.)
We are using the term "vehicle' more narrowly ihan lhe term "medium." For example, a medium might be magazines, but a
vehicle might be a particular knitting magazine
Advertisers ullimately care aboul profits. The optimal advertising strategy will satisfy (a) lhe return on lhe chosen levei of
advertising expenditures is maximized by the choice of media and vehicles and (b) lhe levei ofadverlising expenditures is chosen
ai lhe levei that maximizes profit assuming that lhe choice of media and vehicles is optimal given that levei of advertising
expenditures.
Market Definition in Online Advertising
short time window (e.g., to advertise a short-duration promotion). Magazine ads do not offer that
fiexibility, since the advertiser must commit to a particular message lar in advance.
Television is timely and can be attention getting and effective at generating interest. However, even the
most basic television advertising is relatively costly and is not well suited to communicating iengthy
technicai information. Likewise, newspapers, radio, and outdoor advertising (e.g., billboards) each has its
own set ofstrengths and weaknesses that can be expected to differentially affect that medium's suitability
to a particular advertisers goais.
Because advertising media differ in the degree to which they can target customers, lhe set of consumers
they can deliver to an advertiser differ as well. For example, a skywriting plane cannot target customers
with any more precision than to those who are outside with an unobstructed view of the sky. Television and
radio cannot geographically target customers more preciseiy than a given media market. A biliboard might
target oniy consumers within a neighborhood. An advertisement inside or on lhe exterior of a bus might
target only consumers who reside or work along a particular thoroughfare.
.
Adverlisers may on occasion choose one medium over ali others. However, often an advertiser wili find
it beneficial to select muiliple media; this allows lhe advertiser to target a broader group of consumers and
to utilize their chosen advertising budget oplimally.4' We can expect to see substitution among advertising
media as lhe costs and benefits ofeach of the media vary over time and as the advertising budget responds
to lhe effectiveness of lhe advertising program (the more effective lhe program, the larger lhe advertising
budget).
Furthermore. advertisers also choose within media. To iliustrate, magazines are not monoiithic. Rather,
they differ widely in lhe audience they deliver. lndeed, magazines are often targeled at niche audiences. An
advertiser would seiect the specific magazine or magazines that appropriately target the adverliser's
potentiai customers.
The set of ali television viewers in a Designated Market Arca ("DMA") might be quite diverse (and
therefore not targeted). The set of viewers for a particular station in a particular DMA might be simiiarly
broad or reiativeiy narrow (if it, for example, is a foreign-ianguage station). A particular program on that
station might have a relalively narrow audience and different programs on that station might have very
different audiences, due perhaps both to lhe topic/contenl of the program as well as the time of day it airs.
An advertiser couid choose to run ads during a particular program whose audience best malches the profile
of the advertiser's desired customers.
.
There are, however, limitations on how lar lhe set of consumers receiving ads can be refined. For
example, lhe advertiser could choose a particular television program during which to advertise, bul
typically cannot further selecl within that audience. Anolher advertiser couid choose a particular niche
magazine, a particular issue (e.g., a particular season of lhe year) of that magazine, and even a particular
section of lhe magazine in an attempt to further refine the audience for its ads. in olher media, such as
direct mau, an advertiser may be able to more precisely specify lhe characteristics of lhe audience it would
like to reach.
Advertisers care aboul lhe relurn on lheir advertising doliars. If an advertiser is paying for exposure, the
advertiser would eslimate the expected return on its incremental advertising expenditure on a particular
advertising vehicle by calculaling:
(a) the number ofexposures it expects to receive per doliar of incremental advertising expendilure,
(b) lhe number of incremental saies it expects to receive per exposure, and
(e) its profit, nel of ali appropriale variable cosls, per saie.
' As we explained earlier, lhe optinial advertising budget and the optimal aliocation ofthat budget across media and vehicles are codetennined rather than being separable decisions.
Markei Definition in Online Advertising
The product of (a) x (b) x (e) is the incrementai profit the advertiser expects to receive per advertising
doilar for this choice of vehicle.
Note that component (a) of this caiculation, i.e., the number ofexposures per doilar, isjust the inverse of
the price of the advertising (expressed in do!lars/exposure). Thus the return to a vehicle depends crucially
on the price of advertising on that vehicle, as weli as on the number of saies per exposure (which is likely
to vary across vehicles).42 If this calcuiation (of incremental profit per additional doliar of advertising)
yields a value greater than one doilar, advertising on that vehicle is profitable. (If not, then the advertiser
should not advertise on that vehicle.) Of course, the most pertinent questions are more demanding. First,
which choice of vehicle yields the highest return? Second, what should the advertising budget be?43
For a given advertising budget, a somewhat simpiified analysis would advise that the advertiser spend ali
of its advertising expenditures on the vehicle that yields the highest return. Assurning that each vehicle
choice has a different return, this would imply that the advertiser would spend every advertising doliar on
just one vehicle. This result need not hold, however, when we bring in additional considerations: it may
indeed be optimal for an advertiser to ernploy multiple media and multiple vehicles within media in the
same advertising campaign.
An advertiser may decide to purchase ads on multiple media and vehicles in the sarne campaign for
severa! reasons. First, the advertiser might wish to increase its coverage so as to grow the brand equity of
its products or to increase the value of its trademark. No single advertising medium will reach ali
consumers. Some individuais watch little or no television (or they TiVo through the commercials), while
still being active money-spending consumers. Others listen to their iPods, but not to commerciai radio.
Some read traditional newspapers (i.e., on newsprint) while others read newspapers only online or read no
newspapers at ali.
An advertiser that desired to have a high degree of penetration by its ads into a market would need to
advertise through many different vehicles and media in order to reach a high percentage of the public
through at least one of them. It is quite possible, therefore, that the return per doliar of expenditure on a
single advertising medium may be large up to a point (i.e., for sufficiently small advertising expenditures),
but for sufficient!y large expenditures in that medium, the return to additional expenditures may fali, even
to the point ofunprofitability.
.
Decreasing retums to advertising on a single medium can occur, in particular, when the levei of
advertising approaches saturation of that medium. Decreasing returns can aiso arise because ads may
become iess effective as the number ofexposures ("frequency") increases, users may tire of a particular ad
and stop noticing it ("ad blindness"), and so on. Furthermore, the return to advertising can decline even if
the effectiveness of the advertising itself does not—if the profit per conversion declines. This might occur,
for example, if the advertiser faces increasing marginal costs or capacity constraints.
Second, an advertiser might choose its "media mix" so that potential customers receive complementary
messages. As discussed above, different media have different strengths and weaknesses. An advertiser may
choose to empioy multiple media in order to reach the sarne consumers with messages from multiple
media. For examp!e, an attention-getting poster with little or no hard information at a transit stop may
The incremental profit per saie is unlikeiy to vary by the advertising vehicle. However, there couid be dífferences in the goods a
particular audience buys e.g., the sarne airline ad could generate on different vehicles a different mix of business class and tourist
class licket saies.
' II is inappropriate to view the advertising budget as fixed; if a new media outlet or an irnprovement in the effective targeting
wilhin a particular media outlet increases lhe return on advertising, this should lead to an increase in lhe advertising budget.
1-lowever, the decision-rnaking process of buyers of advertising may not so closely approxirnately optimality. Thus, SiIk, Klein,
and Berndt (2002) inlerpreted the relatively weak own and cross-price elastícities belween different media classes they found for
lhe 1990s as being congruent with a media planning process that "is a sequential one wherein intermedia and intramedia decisions
are separated. Intermedia choices are often effectively preempted by judgments aboul the fit between message strategy and
alternative media exercised in lhe eariy stages of a campaign's deveioprnent and prices are frequently a secondary consideration.
If inteni,edia cornparisons are undertaken at ali, they are likely to be made inforrnally on lhe basis of crileria of uncertain
validity." (Alvin SR, Lisa Klein, and Ernst Berndt, Intermedia Substifutabil/ty and tvfarket Demand &v National Advertisers,
(20:4) REvIEw OF INDUSTRIAL ORGANIZATION 339(2002).
Markei Dejlnition in Online Advertising
complement a television, newspaper, or online ad with more detaiis seen iater.44 Or an online ad can
increase the effectiveness of a television ad.45
This discussion reveais a possibly erroneous impiicit assumption in the traditional analysis of returns on
advertising expenditure. The traditionai analysis implicitiy assumes that the incremental revenue fiowing
from an incremental doilar of advertising expenditure is independent of the levei of expenditures on other
advertising media. in this example, however, expenditures on one medium (the transit-stop poster) wouid
increase the return on expenditures on the other medium. In defining advertising markets, it is essential to
account for the prospect that two or more media may offer compiementary benefits, whether or not they are
economic substitutes.
IV. TYPES OF ONLINE ADVERTISINC
A. Graphic Format
.
Online ads can vaty substantiafly in their graphical format. The simplest ads are text ads, which are fuily
described by a string ofwords, the color and size of the characters, and the dimensions of the ad's bounding
text box. These ads allow only the most basic interactivity: the viewer can ciick on the ad and be
transported to the advertiser's web site or, more specificaily, the ad's "landing page."
A standard "display ad" has more graphical interest. A display ad is typically one of a standard set of
dimensions, with a photograph or other graphic, for example, a logo, and a text message. Like text ads,
such display ads also typicaily allow interactivity through clicking. The graphical element of a display ad is
often anirnated, typicaiIy using Adobe's Flash technology.46
Beyond standard display ads is a wide variety ofmore-complex advertisements, referred to as a ciass as
"rich media" ads.47 Rich-media ads tyically allow for a greater range of interactivity than mereiy a click
which causes the user to leave the original site. lnstead, a rich-media ad can respond to "mouseovers,"
keyboard inputs, or other cIicks that do not result in a click-through to another site. The "footprint" of the
ad can be dynamic rather than fixed: the ad might expand, roll down or roll from the side, peeI back, or
float. Rich-media ads can also incorporate video.
B. Targeting
Online ads also vary in the dimensions on which they can be targeted to various types of customers. As a
result, it is not generaiiy appropriate to say that one type of Internet ad was more or iess targeted than
another. In reality, different types ofonline ads emphasize different consumer characteristics. The result is
that the magnitude of the targeting oftwo different online ads need not be comparable.
.
The decision as to which individuais shouid be targeted for a particular ad involves an advertising
intermediary (which might be a contractual representative of the ad's selier) determining the available
current information about the user visiting the web site. Ifthe intermediary has good information about the
reievant interests of the web site's visitors, the intermediary can choose to display a particular advertiser's
ad if and oniy ifthe current visitor matches the desired profile for that ad.
In order to delve into targeting in more detail, it is useful to distinguish between search ads and nonsearch ads. Search ads are ads that are displayed by a search engine next to the search results in response to
the user having specifled at least one of a set of keywords that the advertiser has identified. Ali other ads
are characterized as "non-search ads."
44
The concept of lhe "marketing funnel" is relevant here: with a marketing funnel consumers are guided via various forms of
- advertising from the initial stages of product awareness and consideration to later stages such as purchase and loyalty.
'° See, for example, Online Publishers Association, "Media Mix Study.' March 2002, ai p. 14. <http://www.onlinepublishers.org/rnedia/ 1 52_W_opa_mediaixstudy_mar02.pdf'
jn
' In 2008, static-graphic display ads represented 39% of graphical ad-serving volume (nieasured by impressions), while simple
Flash-animated ads accounted for 55%. "Rich media" ads, discussed next, accounted for lhe remaining 6% of graphical ad-serving
volume. (AdRelevance, 2008. Cited in DoubleClick and Dynamic Logic, The Brand Value o! Rích Media and Video Ad:, 2 (June
2009), ai http://www.doubleclick.coili/insight/pdfs/The—Brand_Value—of Rich Media and Video Ads.pdf,
Shamim Samadi and Ari Paparo, What's a rich media ad, anvwav?, THE OFFICIAL 0000LE BLOG, (April 30, 2009), ai
http /Igoogleblog.blogspotcomI2009/04/whats-rich-media-ad-anyway.html.
A'íarket DeJ7nition in Online Advertisin,
We consider non-search ads first. When a visitor views a web site's page, the intermediary knows at a
minirnum that the visitor chose to visit that site and that page. (Note that knowledge of lhe particular page
also implies knowledge of lhe information on that page about the lopic discussed.) The amount of
information the advertising intermediary has about lhe visitor mighl be similar to the information that a
magazine has about lhe individuais who read the magazine. Thus, there is an important sense in which a
web site can be like a magazine, newspaper, or radio or lelevision program: the individuais viewing conlent
on a particular site have chosen to view that conlent. That choice is therefore likely to correlate with other
characterislics an adverliser is inleresled in, such as age, gender, income, education, and interests.
In many cases, lhe intermediary will also have useful information about lhe visitor's geographic
localion—perhaps lhe City or metro arca, which can oflen be inferred from lhe visitor's IP address. A
visitor's city or melro arca by itseifwill correlate to some degree with educalion and income, in addition to
specifying lhe visitor's localion 48
.
If a reader registers at a site, lhe seiler of adverlising (or ils intermediary) may have addilional
information about that viewer based on lhe viewer's association with lhe site. If the visitor arrived at lhe
web site from a search engine (i.e., as a resuil ofsearching for one or more keywords on lhe search engine,
which resuiled in a search result for lhe web site, and lhen a ciick on lhe link for lhe web site), lhe web site
will also know lhe search terms the visitor used at the search engine.49
A selier may know more than lhe simple facl of a visil to the currenl page: the selier may know lhe other
pages that lhe visitor visiled that day, or even on previous visils, or whether lhe user clickcd on lhe
adverliser's ad at some prior lime. This is made possible lhrough the use of "cookies," which can be crealed
on a visitor's computer by a web site. Uniess il is deleled or expires, a cookie ailows the web site to
recognize a visitor as using lhe sarne computer as that used by lhe eariier visitor.5°
Knowing lhe seI ofpages the visitor has viewed can give the selier ofadverlising additional information
about lhe visitor's interesls. These inlerests may be hobbies, polilicai altitudes, producls lhe visitor is
considering purchasing, or places lhe visitor may consider as travei destinalions for business or vacation.
Armed with lhis information, lhe seiler of adverlising can display ads chosen because lhe adverliser for
lhose ads is inleresled in adverlising to consurners with characleristics being signaied by lhe visitor's
browsing hislory, localion, and other data.5 '
A selier of search ads will have similar kinds of information as the selier of non-search ads, for severai
reasons. Firsl, lhe search-ad seiler will have information about lhe user's geographic iocation in lhe sarne
circumstances as would lhe seiler of non-search ads (i.e., if a user's IP address can be accurateiy inverted to
determine localion, a search-ad seiler and a non-search ad seiier could both do so). Second, if lhe user is a
regislered user at a farnily of siles that includes lhe search engine, lhe search-ad selier may have
information about lhe user that might be relevanl to the choice of ad to serve, jusl as in lhe case of a nonsearch ad selier. Third, and presumabiy mosl significanliy, lhe selier obviousiy knows lhe keywords lhe
user specified to arrive at lhe search-resuits page. (Recail that in some cases lhe seiler of non-search ads
also knows recenlly specifled keywords.)
To see lhe information, geographical and olherwise, that can be inferred your IP address, browse to http://www.ip21ocation.com/.
The search terrns are encoded into lhe URL of lhe request to the search engine. This URL is passed along to lhe destínation web
site as lhe referrer fleld ofthe header of lhe HTTP request to the destinalion site.
When consurners register at a web site, the site typically has access to information about the consurner's browsing habits on that
site. That inforrnation can be expecled to be at leasl as reliabie as that provided by cookies in the absence of registration. For
example, a user regislered at nylinies.com is recognized as that person whether lhe person v,sits the site from her borne computer,
her iPhone, her work computer, or the computer at a Business Cenler in the hotel where she stays on a business Irip. In contrast,
cookies would view an unregistered individual's visits from two different computers as if she were two differenl individuais.
Conversely, two differenl unregistered individuais browsing under lhe sarne user accounl on lhe sarne computer would
erroneously be inlerpreted as being lhe sarne person.
" According to Yahoo!, its behavioral-targeting capability "goes beyond lhe more common rules-based segmenlalion or cluslering
of users by siles visited" to include as well searches and ad inleraclions. Yahoo!'s "sophislicaled modeling technology" lhen
predicls where a user sits in lhe "Awareness-Consideralion-Purchase funnel" in order to deliver lhe appropriale adverlising
message. Yahoo! Advertising, Behavioral Targezing, at htlp://adverlising.yahoo,cornladsol ulion#productBehaviorai.
lo
Market Definition in Online Advertising
Depending on the particular search engine and on the particular user, the seiler of search ads may have
additional information, as in the case of the seller of non-search ads. With respect to search ads, search
engines can be expected to know other search terms that the user has specified in at least the recent past. As
we have discussed, Google is beta testing targeting ads based on sites and pages visited over some time
period, rather than only the current page being visited. It might also be possible for Google to merge these
sources of information about a user to provide a fuller perspective on the user's interest than could be
gleaned from conteniporaneous search behavior alorie.
C. Basis of Payment
In most offline advertising, it is very difficult to track a consumer's reaction to an advertisement.52 As a
result, advertisers pay for radio, television, newspaper, magazine, and biliboard advertisements on the basis
ofconsumers' anticipatcd exposure to the ad rather than their actions as a result of seeing the ads.53
In contrast, lhe inherent interactivity allowed by the hypertext nature of web pages creates the possibility
of measuring, and basing the advertiser's payment upon, responses by the visitor; specifically, whether the
visitor clicks on the displayed ad. As a result, there is the option in online advertising for advertisers to
"pay per click" rather than solely paid by impression .54 lndeed, search ads are typically sold on a per-click
basis, whereas banner display ads are often sold on a per-impression basis.
However, display ads need not be sold on a purely per-impression basis. Display ads are typically
clickable (so that clicking on the ad takes you to the advertiser's web site) and some display ads
aggressively encourage such interaction. It is quite possible for display ad charges to be made on a hybrid
basis, paying both per impression and per click. lndeed, Google's AdWords customers can purchase ads to
be displayed on lhe web sites ofany ofGoogle's AdSense partners, and these display ads are charged either
on a per-click or per-impression basis.55
V. ADVERTISING COMPETITION
A. Defining Relevant Markets
The choice of a relevant market is typically the first decision that is made in the antitrust review of a
proposed acquisition or in the litigation ofantitrust claims. It is important to note that this question cannot
be meaningfully asked in the abstract. Defining a relevant market should be done with respect to a
particular acquisition or with respect to specitic challenged conduct.
.
For example, in a case of alleged monopolization of a local market for newspaper advertising, it will be
important to understand the degree to which online advertising competes with, and disciplines the prices of,
those local newspapers' advertising. Similarly, in a case alleging monopoly power in some Internet-based
advertising market, it would be important to understand the degree to which offline advertising competes
with, and discipline the prices of, online advertising.
The answers to these two questions need not be the same. It is quite possible that that online advertising
disciplines the exercise of market power by sellers ofoffline advertising in one context, but that in another
context offline advertising would not discipline the exercise of market power by sellers of online
advertising. 01' course, the converse could be true in other situations.56 Whatever the context in which the
relevant market question is asked, however, the answer requires an understanding of the nature and extent
ofthe competition between different forms of advertising.
The effectiveness of some offline advertising can be measured. For example, a manufacturer can monitor how many of its
coupons are redeemed. A nierchant can advertise "mention this ad to receive your discount."
A department-store merchant, John Wanamaker (1838-1922), is credited with having said: "Haif the money 1 spend on
54
advertising is wasted; the trouble is 1 don't know which haif."
Note that paying per impression for an Internet ad may carry less uncertainty than paying per impression 0ff une. If an advertiser
pays for a display ad in a newspaper, lhe advertiser pays based on the newspaper's circulation and anticipated newsstand saies.
However, the advertiser will not know on a gven day Iiow many ofthose subscribers or newsstand purchasers actually open their
papera to the page on which the advertiser's ad is Iocated. When a page view is served on the Internet, however, it is highly likely
that lhe consumer explicitly requested it in order to read it.
55 See Google, Websíte Advertising, aí http://www.goog1e.com/advertisers/onIine/website.htm1.
v In other words, for the purposes of niarket definition, the relation "competes with" need not be reflexive.
32
11
Ivíar/cei Definition in Online Advertising
We will discuss lhe specifics of market definition in the sections that follow. We note that the U.S.
DOJ/FTC Horizontal Merger Guidelines provides a useful framework for the analysis ofrelevant markets.57
The Guidelines begin wilh a demand-side analysis, asking whether a monopolist of a hypolhesized product
market could profitabiy raise prices a smaii but significant arnount above the "competitive levei" for a
significant period of time. if the answer is yes (and the conjectured market is the srnailest such market for
which lhe answer is yes), the hypothesized market is a reievant market; if the answer is no, the market is
expanded in its breadth ofproducts and lhe exercise continues. Supply side queslions relating to entry and
competilive responses such as repositioning are treated primariiy as informing questions of market power
rather than market definition.
It is not aiways possible in practice to undertake a complete Guidelines analysis of this type for ali types
of advertising. Neverlheless, the framework poses lhe right question when mergers and acquisitions are
invoived. In the non-merger context, the applicability of lhe Guidelines is less clear. Thcre are at ieast two
reasons for this. First, if an alleged anlicompetilive practice invoives a firm with substantial market power,
currenl prices may not be competitive.58 Second, lhe smai!est reievant market may not be the market of
interest if the practice at ieast invoives a broader set ofcompetitive products or services.
.
B. Competition between Offline and Oniine Advertising
Online and offline advertising serve the sarne broad advertising goals. Advertisers are typicaily trying to
inform, persuade, remind, or motivale consumers by delivering informalion, rheloric, and/or imagery to
those consumers. This is true whether lhe adverliser chooses to advertise offline, online, or both. Oniine
advertising is capabie of accompiishing lhese goals jusl as is offline adverti sing—keeping in mmd that
every medium and vehicle has its own peculiar set ofstrengths and weaknesses.
1. Does Online Advertising Competitively Constrain Offline Adverlising?
if one had looked at competition for offline ads a decade ago or perhaps even five years ago, online
advertising would not have provided a significant competitive threal. lndeed, before lhe risc of the Internet,
anyone but a hermil was reachabie by offline advertising and unreachable by the not-yel-bom medium of
online advertising.
However, as the Internet deveioped and consumers started spending time online, the audience for online
ads deveioped. Ullimaleiy, lhe irnportant question wilh respecl to whether online advertising disciplines the
pricing of offline ads is whether advertisers would be wiliing to shift sufficient advertising expenditures
from offline to online in response to a hypothelical across-the-board increase in the prices of offline ads to
render such price increases unprofitable.
The answer to this "criticai ioss" question is informed but not answered by informalion aboul lhe growth
of Internet advertising. There is no doubt that in lhe past Vive years the growlh of lhe Internet has made
online advertising more of a competitor to offline advertising. Yet, it remains lhe case now and for at least
severai or more years that online advertising represents and wiii represent oniy a smaii part of total
advertising spend. Aggregated into broad categories, lhe iargesl part of total US advertising spend (37.2%)
in 2008 wenl to print.59 The next iargesl part (30.3%) went to TV advertising.60 in 2008, online advertising
accounted for only 10.6% ofaii US advertising spend»
57 U.S. Depariment of Justice and the Federal Trade Comrnission, Horizontal Merger Guidelines, (April 8, 1997),
- http:l/www.usdoj.gov/atr/public/guidelineslhmg.htm.
See discussions of lhe associated "cellophane fallacy" in, for example, Jonathan Baker, Markel Dejlnition: An Analytical
Overvie,v, (74:1) ANTITRUST LAw JOURNAL, 62-165 (2007) and Lawrence J. White, Markei Power and Market DeJ'inhlíon in
Monopolization Cases: A Paradigrn is Missing, ISSUES IN COMPETITION LAW AND POLICY (Wayne D. Collins, ed., 2008),
http://www.usdoj.gov/atr/public/hearings/singlefinii/docs/222 1 04.pdf.
1 lere, lhe print category contains newspapers, custom publishing, consumer magazines, and 13213 magazines, but not Yellow
Pages
Here TV consists of local and national spot TV, cable network TV, broadcast network TV, local and regional cable TV, and
broadcast syndication TV.
eMarkeler, Online Ad Spending Siows bui Grabs Marker Share,
(September 21, 2009),
ai
http://www.emarketer.coiiVArticle.aspx?R=1007283. The data ciled are from the Jack Myers Media Business Report,
"Advertising & Marketing lnvestment Forecast 1998-2012."
12
1Market Dejlnhtion in Online Adverlising
The sarne source forecasts the shares of total advertising spend of the various types of advertising
through 2012. The rate of increase of online advertising's share of total US advertising spend is forecast to
slow dramatically. The rate ofgrowth of online advertising's share from 2007-2008 was alrnost 18%.62 By
2012, in contrast, the annual rate of change of online advertising's share is forecast to be less than 2%,
reaching 13.6% by 2012.
Whether online advertising is sufficiently competitive to discipline offline ad pricing generally is an
open question, whose answer will almost certainly change over time.
A look at various types of offline advertising yields a number of interesting questions. it is plausible to
think that online advertising, coupled with other forms of offline advertising, might discipline the exercise
ofmarket power in some particular form of offline advertising that is alleged to be a relevant market.
•
Looking more specifically at one particular subset of offline advertising—newspaper advertisingprovides a timely and pertinent exarnple.63 The decline of newspapers has been prominently in the news.
Newspaper Death Watch keeps a running tally of US metropolitan dailies that have closed since Death
Watch began publication in March 2007. That Iist includes papers such as the Baltimore Examiner, the
Rocky Mountain News, and the Tucson Citizen. There are also newspapers that have ceased online
publication or moved to a hybrid online/print modei. Examples of these include the Seattle PostIntelligencer, the Detroit News/Free Press, and the Christian Science Monitor.64
From 2007 to 2012, newspapers' share of total advertising spend is forecast to drop 6.6 percentage
points, from 18.7% to 12.10/o—a 35% decrease in share.65 This decline in newspaper ads has been
precipitous, and does not characterize offline advertising generally. Over the period 2007-2012,
newspapers' share of total US ad spend is forecast to decline more quickly than online advertising's share
of spend is forecast to increase, showing that non–newspaper offline advertising is also increasing its share
of total advertising spend.66
It is not clear, however, that merely observing a shift in advertising expenditures from newspaper
advertising to online advertising is sufficient to support the argument that online advertising competes with
and disciplines the prices of newspaper advertising. The confounding fact is that, not only is newspaper
advertising declining, but newspaper readership is declining as well. For example, daily print newspaper
circulation is down 13.5% from 2001 to 2008. (Sunday circulation is down 17.3% over the sarne period. )67
Weekday circulation averaged over 395 daily newspapers was recently reported to be down 7.1% relative
to a year prior.68
Thus the eyes that advertisers need are themselves moving from newspapers to online content. It would
be natural for advertisers to follow the flow of eyeballs with their own advertising expenditures. The
question ofwhether online advertising disciplines newspaper advertising prices is more narrow and, to our
knowledge, not resolved: would an increase in newspaper-advertising prices above the competitive levei be
02
To avoid any rnisunderstanding, this means #hat the share as a percent changed by alrnost 18%, not by 18 percentage poinis. In
2007, online advertising represented 9.0% of total US advertising spend; in 2008, it represented 10.6% of total US advertising
spend. Thus from 2007 to 2008, online advertising's share of total US advertising spend increased by alrnost 18% (10.6%9.0%)/9.0%.
03
When we reler to newspaper advertising or readership, we are referring to advertising in, or readership of, the traditional
newsprint editions of newspapers; we are excluding newspapers' web editions.
64
Newspaper Death Walch, ai http://www.newspaperdeathwatch.com/.
eMarketer,
Online Ad Spending Siows but Grabs Markei Share,
(September 21, 2009),
ai
http://www.emarketer.com!Article.aspx?R=1007283. The data cited are from lhe Jack Myers Media Business Report,
"Advertising& Marketing Invesirnenl Forecast 1998-2012."
Over that sarne period that newspaper advertising is forecast to lose 6.6 percentage points, the share of online advertising is
expected to increase by only 4.6 percentage points, from 9.0% to 13.6%. Therefore other offline advertising categories are gaining
aI newspapers' expense.
Pew Project for Excellence in .lournalisrn, The Staie 0/ the News Media 2009, citing data from lhe Newspaper Association of
Arnerica
through
2007
and
adjusting
it
ai
for
further
losses
in
2008,
http:/lwww.stateoflhernedia.orgI2009/narrativenewspapersaudience.php.
00
Tim Arango, Fali in Newspaper Sales Accelerates to Pass 7%, THE New YORK TIMES, (April 27, 2009), ai
http://www.nytirnes.corn/2009104/28/businesslrnedia/28paper.htrnl. The figures reflect lhe six-rnonth period ending March 31.
2009, compareci to lhe sarne period a year prior.
13
Markei Dfinition in Online Advertising
unprofitable as a resuit of advertisers' consequentiy more-rapid defection to online advertising vehicies?
Given consurners' shift from newspapers to online, some part of the corresponding shift of advertising
cannot be attributed to price competition.
Notwithstanding the above uncertainties, courts have recognized the existence of competition that the
online world poses the offline one. For example, in a hotly contested case invoiving the acquisition of the
San Francisco Chronicle newspaper by the Hearst Corporation, the Court cornmented in dieta that, from
newspaper readers' perspectives, there are many other information sources, inciuding online sources that
a more recent case invoiving the Gazette Newspapers, the aileged newspaper advertising
compete.6')
market was deemed to be too narrow because it improperiy excluded other forms of print and non-print
media advertising.7°
2.
Does Offline Advertising Competitive/y Consírain On/ine Advertising?
A separate question is the extent to which offline advertising continues to be a competitive constraint on
most or ali forrns of online advertising. It is quite possible that offline ads wili constrain some, but not ali,
online ads. For some advertisers and for some consumers, offline advertising may not be a competitive
option. To make sense ofthis issue, we need to evaluate the specific objectives of the advertisers (e.g., their
target audience) and we need to estimate the number of "marginal advertisers" who would find it profitabie
to increase their reiiance on offline ads ifthe price of online advertising were to increase (the "actual ioss").
Depending on the margins generated by the advertising, only a relatively small number of advertisers may
need to be wiiling to switch (the "critical ioss") to make a price increase unprofitable. if the actual number
who would switch is greater than those, it would be appropriate to put offline and online advertising ofthis
particular type in the sarne relevant market.71
We believe that both online and offline advertising are iikeiy to remain robust into the foreseeable
future, suggesting that the two wili compete quite generaliy. Offline advertising is diverse in the very many
different media it involves, everything from direct mau, to television ads, to prornotional swag, to
illuminated dirigibles. There is no reason that—as a ciass—offline advertising will become unimportant in
lhe foreseeable future notwithstanding the increased (and inereasing) leveis of online activity. For that
reason online and offline advertising can be expected to compete activeiy for advertisers' doiiars for the
foreseeabi e future.72
The decline of the American newspaper discussed previously raises the question of whether this
particular offline advertising medium wili continue to discipline online advertising. Although offline
adverti sing—largely because it subsumes television and radio adverti si ng—shows no sign of going away,
the sarne cannot be said for the American newspaper. As traditional hardcopy newspaper readers shift to
reading news on the Web (whether from online versions of traditional newspapers or from online-oniy
sources such as the Huffington Post) and as newspapers continue to dose, it would be reasonable to
Reillv v. Hearsf Corp., 107 F. Supp. 2d 1192 (N.D. Cal. 2000). Note, however, that the plaintiff was not alieging an advertising
market but ralhei- a market for the provision of"daily newspaper news, features and opinion." Thus Judge Waiker's dicta did not
speak directly to whether online advertising competes with newspaper adverlising.
Berlyn Inc. v. Gazetie Newspapers, 73 Fed. Appx. 576, 582-83 (4th Cir. 2003). The Couri also concluded that lhe alleged relevant
rnarket was 100 broad because II cornbined legal advertising (i.e., notices) and comrnercial advertising in lhe sarne market.
n However, these advertisers mighl still be protected by lhe competition between offline and online advertising if selters of
advertising could not identify and price discriminate againsl these advertisers' ads aimed aI these discrete offline-oniy customers.
In its investigation of Google's acquisition of DoubleClick lhe European Comniission declined to define a broad niarket for ali
advertising media in which lhe "internet would be jusl one of lhe several media channels—among which TV, newspapers, etc.that can be chosen by advertisers wanting to promote their goods or services." The EC so found "primarily because lhe market
investigation revealed that offline and online advertising are perceived as separate markets by lhe majority of respondents."
Additionally, lhe EC cites lhat "online advertising is used for specific purposes," can be better targeled, "has a unique reporting
system" to belter measure ad effecliveness, and has unique advantages of both lhe pay-per-irnpression and pay-per-click payment
niechanisrns. (Comrnission of lhe European Communities Decision of 11/03/2008 declaring a concentration to be compatible with
lhe cornmon market and lhe functioning of lhe EEA Agreement. Case No COMPIM.4731 - Google/DoubieClick, Regulation (EC)
No 139/2004 Merger Procedure, Arlicle 8(1). C(2008) 927 final. Public Version. at ¶ 44-46) We note that these arguments by lhe
EC suffer lhe sarne basic problem as the argurnents of the FTC we discuss below: Merely identifying differences is characteristics
between two producls does not conslitute a valid argument for placing them in separate relevant product markels. The EC leaves
lhe central question unanswered: whether offline and online advertising are sufficienlly dose economic substituIes lhal each
disciplines lhe prices ofthe other.
14
Market Definilion in On/ine Adverlising
conclude that offline newspaper advertising will decline in importance as a possible competitive constraint
on online advertising.
This possible effect on competition is worth mentioning—even though there are many other sources of
offline adverti sing—because offline newspaper advertising may be lhe closest substitute for online
advertising of ali the offline advertising media. Offline newspaper advertising would appear to be very
similar to text and non-video graphic display ads at web sites (particularly so at online newspaper web
sites). Furthermore, offline newspaper advertising may be best suited among offline advertising media to
deliver the sarne type of inforrnation to the sarne type of consumers as does online advertising.
C. Competition between Online Search and Online Non-Search Advertising
In this section, we assume for purposes of discussion that some or ali forms of online advertising are not
disciplined by offline advertising. In other words, we assume that the relevant antitrust market or markets
are no broader than ali online advertising. We now ask whether online search and non-search advertising
are sufficiently separable so as to be placed in separate markets, or whether they are sufficiently
competitive with each other so as to be seen as representing one market.
We have discussed both similarities and distinclions between search and non-search ads. The
characteristics-based analysis we have described thus far is by itself not clearly determinative with respect
to market definition.
The best evidence might come from historical studies of past quasi-experiments in which there was
variation in the relative price of search versus non-search advertising. Because of the rapid evolution of the
online advertising industry, this exercise would be a difficult one. Indeed, we are aware of little direct
econometric evidence pertaining to this important question.
One contribution is an econometric study by Avi Goldfarb and Catherine Tucker of a "natural
experiment" that arises from variation across states in regulations limiting whether iawyers can contact
clients by mau.73 Forbidding marketing via mail can be interpreted as an artificial barrier to any
competition that otherwise exists between offline (e.g., snail mau) and online advertising. interestingly, the
authors found that in areas where lawyers could not market by mail prices were on average 7 percent more
for relevant search-ad keywords such as "tax lawyer." The authors' interpretation of these results is that
competition from offline advertising disciplined the price ofonline advertising, reducing the price ofonline
advertising by 7 percent; they say their results suggest "that online context- advertising competes in a
broader advertising market that includes offline marketing communications channels."74
.
However, the scenario these authors study is extreme and therefore ofdoubtful relevance to deterrnining
the boundaries of a relevant market. Market definition is typically conceptualized by considering a modest
price increase (e.g., 5 percent) and whether there is substitution driven by the change in relative price. A
prohibition of snail-mail marketing is, in contrast, equivalent to an infinite (or at least a very large) price
increase. That such a large price increase might cause substitution from offline to online says little or
nothing about whether a small price ncrease typically used in relevant-market analysis would lead to
substitution.
It is relevant to review the analysis of the Federal Trade Commission ("FTC") in its evaluation of lhe
relatively recent Google-DoubleClick acquisition. In closing its investigation of the acquisition, the FTC
found that there could not be a relevant product market that contained both search and non-search
advcrtising.75 According to the FTC:
[T]he evidence in this case shows that the advertising space sold by search engines is
not a substitute for space sold directly or indirectly by publishers or vice versa. Or, to
Avi Goldfarb and Calherine Tucker, Search Engine Advertising: Pricing Ads to Context, NET INSTITUTE WORKING PAPER 407-23
(September 2007), (hereafter "Goldfarb and Tucker (2007)"), http://ssrn,comlabstract= 102145 1.
Goldfarb and Tucker (2007) ai p. 18.
Federal Trade Commission, Staterneni of Federal Trade Conirnission Concerning Google/DoubleClick, (FTC File No. 071-0170)
3 and 7 (December 20, 2007), (hereafter "FTC (2007)"), http://www.ftc.gov/os/caselist/0710170/071220statement.pdf.
15
tvíarket Definilion in Online Adverlising
put it in terms of merger analysis, the evidence shows that the saie of search
advertising does not operate as a significant constraint on the prices or quality of other
online advertising sold directiy or indirectly by pubiishers or vice versa.
A review of the factual basis and reasoning behind the FTC's finding wiii be instructive.
1.
The FTC claims that search engine marketing is ~for branding
According to the FTC,
Businesses purchase search ads for different purposes than businesses purchase nonsearch ads and "one type does not significantiy constrain the pricing of another"
"For instance, advertisers primarily purchase search advertising space to impiement
direct response ad campaigns, whiie directiy soid ad inventory is generaiiy purchased
for brand advertising campaigns."76
.
The FTC repeats a common stereotype of the search vs. non-search advertising distinction: that nonsearch dispiay advertising is for branding and that search advertising is not. Howevcr, market research
chalienges this claim. The most-recent State of Search Engine Marketing survey, conducted by SEMPO
(the nonprofit Search Engine Marketing Professional Organization) gathered detailed responses on many
topics from 890 search engine advertisers and search engine marketing (SEM) agencies.77
Among other questions, the survey asked "What is your company using search engine marketing to
accompiish?" There were six possible responses, pius "other;" muitipie responses were aiiowed.78 Iii
contrast to the FTC's position, a supermajority (63%) of ali respondents stated that they used search engine
marketing "[tio increase/enhance brand awareness." Among respondents with greater than 500 staff, 70%
rcsponded they were using search engine marketing in order to increase/enhance brand awareness.
Moreover, this brand-buiiding response received lhe mosi rnenhions out of ali seven possible answers.79
This result is consistent with an IAB/Nieisen study demonstrating "that sponsored text advertising in the
search environment works for an array of branding objectives." Sponsored text ads had the biggest impact
on the "unaided brand awareness" metric, where the strength of the effect was reiated to the prominence of
the position of the text ad.80 Of course, this dynamic works in the other direction as weli—not oniy are
search ads increasingiy being used for branding purposes, but advertisers are also becoming aware of the
ability to use dispiay for direct-response campaigns.
2.
The FTC claims thai lhe uniqueness ofsearch engines 'characleristics pul search in a separale
relevam rnarkeifrorn non-search advertising
The FTC points to characteristics of search ads:
"[S]earch engines provide a unique opportunity for advertisers to reach potential
customers" and "users visiting a content page do not declare their interests in the sarne
way they do when they type in a keyword on a search engine."t '
We have pointed out previousiy that the intermediary that decides what ads to serve to which user on
which web page has different information availabie to it when serving a search ad than when serving a nonsearch ad. As we pointed out, search and non-search ads are differently targeted. However, it cannot be said
that one is more or iess targeted than the other.
' FTC (2007) at p. 7.
SEMPO, Executive Sumrnary, The Siate of Search Engine Marketing 2008, (February 2009), (hereafter "SEMPO (2009)"),
http://www.sempo.org/Iearning_center/research/2008_execsummary.pdf.
The other allowed responses were "[t]o seIl products, services or content directly onhine," "[t]o generate Ieads that we ourselves
wihl dose," "[tio drive traffic to our ad-supported web site," "[tio generate Ieads for a dealer [or] distributor network," and "[tio
provide informationah/educationah content onhy."
SEMPO (2009), at slide 8.
° IAB and Nielsen/fNetRatings, IAS Issues New Research Fron Nielsenl/Netrallngs On Branding Vah,e O! Sponsored Texi
Advertising, (July 15, 2004), ai http://www.iab.net/about_the_iab/recentpressreieases/press_reieasearchive/pressrelease/4742.
' FTC (2007) at p. 3. [emphasis added]
16
Markei Dejlnition in On/ine Advertising
Based on the pubiicly availabie evidence cited by the FTC, their conclusion that search and non-search
do not compete is not compeliing. In its essence, the FTC is suggesting that the two classes of ads do not
compete because they have different characteristics and in particular are differentially targeted. Fiowever,
the ultimate market definition question depends on whether the two products are sufficiently dose
economic substitutes so that each constrains the pricing of the other. This central question remains
unanswered.
3. The FTC clairns 1/ial pay-per-click and pay-per-irnpression advertisements cannot be in lhe sarne
re/evan! market
The FTC argues that contextuaily targeted ads sold through intermediaries are not substitutes for directly
purchased display ads. The FTC reaches this conclusion not only because the ads are sold through two
different channels, but also seemingly because the FTC believes ads sold on a pay-per-click basis cannot
discipline the prices of ads sold on a pay-per-impression basis, and vice versa.82
This argument is not compelling, for two reasons. First, advertisers can weigh the tradeoffs between an
ad sold on a pay-per-click basis and an ad sold on a pay-per-impression basis in each case, the advertiser
can estimate the cost of running the ad and decide which of the two ads has the better prospect.85' 54
Second. the FTC does assert that contextual and non-contextual ads—when sold by an interrnediaryare in a single relevant rnarket:55
However, the evidence shows that the prices and quaiity of contextual ads are
constrained by other forms of display ads sold by ad intermediaries (and vice versa).
We therefore determined that contextuaily targeted ads do not constitute a separate
market rather they are part of a broad market that includes ali ads sold by
intermediaries.
We have no reason to disagree with the FTC's conclusion with respect to intermediary-sold ads. We
simply note that contextual ads are pay-per-click and non-contextual display ads are not. We agree with the
point that is impiicit in the FTC's analysis: the fact that the two types of ads currentiy rely on different
pricing mechanisms is in itself not supportive of the conclusion that search and directly sold display ads are
not in the sarne market.56
As is likely the case with offline advertising, there will aimost certainiy be some convergence over time
between the types of search and non-search ads. The result will be that some of the differences commoniy
associated with search vs. non-search ads will become inessential distinctions, ifthey are not so aiready.
For exampie, search ads today usually involve text and not graphics, whereas many non-search display
ads invoive graphics. Graphicaliy rich display ads, even video, could be dispiayed above, below, and next
to search results. The choice to use only text-only ads on search-resuits pages might be motivated by the
desire ofsearch-engine companies to keep the response speed of their search engine as high as possible. As
broadband speeds increase, and as high-speed broadband Internet access achieves greater penetration,
richer content will likely begin to be displayed as search ads. Yahoo! has tested and initiated a iimited
52
FTC (2007) ai p. 5.
° A cosi per click can be translated into a cost per impression, and vice versa, by first estimating the ad's click-through rate. lndeed,
adverlisers make such a calculation when buying AdSense ads. We understand that Google also seus search ads using cost-perimpression pricing in China, since lhe compiemenlary factors for pay-per-action advertising (credit cards and package deuiveiy)
are not as niature as in some other countries.
We note, however, that lhe need for a trade off here only arises if there is some kind of advertising budget. Oiherwise lhe change
in lhe return of ad opportunity A shouid foi change how much one spends on ad opportunity 8 (as long as lhe price of ad
opportunity B doesn't change and the conversion rate of ad opportunity Bis foi affected by changes in lhe levei ofspend on ad
opportunity A. We previously discussed factors that couid give rise to lhe existence of a budget, such as increasing costs or
capacity cOnstraints.
FTC (2007) ai pp. 5-6.
ii also noteworthy that pay-per-click and pay•per-impression ads compete in lhe sarne ad auctions in Google's AdSense program.
17
Markei Definition in Online Advertising
rollout of its Rich Ads in Search program "that lets advertisers add video, images and custom search boxes
to their search ads.'87
Today, many non-search ads are text-only like search ads. For example, ads served on Google's
AdSense nctworks are not search ads—they appcar on traditional web sites—but they are primarily textonly. The salient distinction seems to be that search ads are targeted on the basis of the consumer's mostrecent search request.
It is imporlant to ask about the competitive significance of the difference between search and non-search
ads. Search ads and non-search ads are both targeted, but with different emphases as we explained above.
While the targeting dimensions are different, there are many similarities between the information about
visitors held by search-ad sellers and the information held by non-search ad sellers. Both have equal access
to geographic information about the visitor. Both may have additional information about a visitor if the
visitor is registered at the site. Both may have information about some of the recent browsing behavior of
the visitor, depending on the breadth of the site visited and on the site's policies.
.
Sellers of search ads have information about the search query the consumer is conducting literaily right
that moment. Seliers of non-search ads have information about the content of the web page the visitor
chose to view and is viewing that moment. Further, sellers of non-search ads would know the search
keywords the visitor used in the case that the visitor arrived at the site via a search at a search engine.
Therefore sellers of both types of ad inventory have some information about the visitor, the visitor's
browsing, and about products, activities, or topics the visitor may be interested in as signaled by her
browsing and search activities.
Although there is some overlap between these two sets of types of information, there are also
distinctions. These dislinctions are in different dimensions—non-search ad sellers may have more
information about what a visitor's general interests over time are, whereas a search-ad selier may know
more about what is on the visitor's mmd at a moment when the visitor is searching for something in a
focused way—as opposed to simply being a more precise estimate of some customer characteristic.
Therefore it cannot be sensibly said that either search or non-search ads are "more targeted" than the other.
Thcy are
d(Jjerent/y targeted.
It may be that—for a particular advertiser—one type of information about visitors betler corresponds to
the advertiser understanding of her own customers. That customer might then in some sense prefer either
non-search over search ads or search ads over non-search ads. The qualifier might... in some sense is
necessary because one cannot meaningfully discuss an advertiser's preferences between two types of
advertising without also speciving the prices of lhe advertising types.
Note that two types of ads can compete even though they are differently targeted. Even if it were the
case that one type (search or non-search) was unambiguously better targeted than the other, this would not
imply that the two types of advertising are not dose substitutes or do not compete. In fact, in such a case
search and non-search can strongly price compete because the price of non-search ads can adjust to make
the advertiser indifferent between the two types of ads.
To see this, we íirst compare (a) a pay-per-click search ad vs. (b) a pay-per-click non-search ad, where
we assume that the non-search ad is targeted toward the advertiser's customers only half as well as is lhe
search ad (i.e., it takes twice as many impressions of the non-search ad to generate a single click as the
number of impressions of the search ad required to generate a single click). As we have discussed earlier,
because the advertiser is paying per click, lhe advertiser is no worse offbuying the non-search ad as buying
the search ad. It is lhe seller of the advertising space (e.g., lhe publisher of the content web site) who is
unable to monetize its inventory as well as the search site is, because—according to our assumplion—lhe
search site is able to belter target its ads. Thus, pay-per-click non-search ads can strongly price compete
with search ads.
Jeff Sweat,
Your Ads, Richer,
YAHOOÇ SEÂRCH MARKETING BLOG.
http:/fwww.ysmblog.corn/blog/2009/02/ 1 8/your-ads-richer/.
18
(February 18, 2009),
ai
Markei Dejlnition iii On/ine Advertising
Now we compare (a) a pay-per-click search ad vs. (b) a pay-per-impression non-search ad. Recail the
eariier point that advertisers u!timate!y care about the return on their advertising expenditure. The rate-ofreturn calculation is different depending on whether the advertiser pays per c!ick or per impression. The
return for advertisers that pay per impression (such as many non—search ad advertisers) is a function of
(a) lhe price per impression, caii thisp1;
(b) the impression-conversion rate (the rate at which visitors viewing the ad actualiy purchase the
product), ca!! this K1 and
(e) the incremental profit from an incremental sa!e, cai! this 2r.
However, the return for advertisers that pay per click (such as search advertisers) is a function of
(a) the price per c!ick, caii lhisp';
(b) the ciick-conversion rate (the rate at which visitors c!icking on the ad actua!iy purchase lhe
product), ca!! this K; and
(c) lhe incremental profit from an incrementai sa!e; this is the same ,rin the per-impression case.
Suppose lhe advertiser spends B on lhe non-search pay-per-impression ad and B on lhe search pay-perc!ick ad. The resu!ling gross profit for the advertiser wi!i be, for the non-search ad:88
-B. K1 • ir,
PI
and for the search ad:
B . • 2T•
-'K(
PC
The gross protils from spending B doi!ars on lhe non-search ad wiii equal the gross profits from
spending 8 do!iars on lhe search ad if:
Pi
K
1
i.e., if lhe ratio of price per impression to price per c!ick is lhe sarne as lhe ratio of the conversion rale per
impression to the conversion rale per c!ick. For examp!e, if lhe number of saies per impression of lhe nonsearch ad were one-lenth of the number of saies per c!ick of the search ad, lhen the per-impression price for
the non-search ad that wouid make the advertiser indifferenl between the non-search ad and lhe search ad
wouid be one-tenth of the per-ciick price for lhe search ad.
At ils most fundamental, an advertiser buys saies. The advertiser on!y cares about how rnuch she pays
for each saie. She does foI at the end of lhe day care whether she paid for the sa!es on a per-ciick or a perimpression basis or whether it required many impressions or on!y a few. If lhe non-search ad is iess weli
largeted lhan lhe search ad, lhe implication is that lhe advertiser requires a !arger number of impressions of
the non-search ad to achieve lhe sarne saies as wilh the search ad. The advertiser can achieve the sarne leve!
88
The first factor (Bip,) is the number ofimpressions that B doliars can purchase. The second factor is the number of saies that result
froni each impression. The product of lhe first two factors is the number of saies that resuit from spending B dollars on the nonsearch ad. Multipiying by ,ryields lhe gross profit from spending 8 doliars on the non-search ad.
19
Market Dejlnition in Online Advertising
NEW
of saies using the non-search ad and at the sarne cost as long as the non-search ad's price adjusts to
compensate for its iower specificity oftargeting.899°
it has been suggested that the fact that the prices of search ads are set by auction removes any
competitive concern. Google co-founder Larry Page, in the context of explaining why advertisers would
benefit from a proposed Google-Yahoo advertising partnership, said:9 '
AdWords is an auction. We're not setting prices. Auctions are determined by suppiy
and demand.
Simiiariy, the Economist has asserted that "[o]n antitrust, the price that Google charges its advertisers is
set by auction. so its monopoiistic clout is iimited."92
There is no doubt that the auction process does add an important competitive element to the advertising
business. However, auctions are not a cure-ali for concerns about exercise of market power. in general,
auctions determine a price that ciears suppiy and demand. For a given demand, the iower the supply the
greater wiil be that market-ciearing price. Auctions move any concern about the exercise of market power
from a concern about the direct setting of prices to a concern about a restriction of output that couid have
the effect of raising prices.
Suppose it was the case that search advertising is not effectiveiy disciplined by non-search advertising.93
Wouid the auction mechanism itself prevent a hypothetical monopoiist of search advertising from raising
search-ad prices above competitive leveis? First, such a hypothetical monopoiist would have some
discretion over the total quantity of search-ad inventory offered for saie. By, for example, reducing the
number of search-resuits pages on which it soid ads or by reducing the number of ads per page, the
hypothetical monopoiist couid raise the prices set by the auctions.94 Furthermore, the hypothetical
monopolist couid set minimum bids, which couid also have the effect of raising auction prices
general iy.95'96
In an auction for a particular search term, the winning bidder pays a price reiated to the bid of the nexthighest bidder rather than to the winner's own bid. Thus within a search term the auction cannot fuiiy price
discriminate by charging an advertiser more as a result of that advertiser having a higher vaiue for the ad.
However, the partitioning of search-ad piacements into hundreds or thousands of keyword groupings,
followed by the separate auctions and separate resuiting prices for these keyword groupings, creates the
89
•
The judge in KinderStart v. Google rejected piaintiff's claimed search-ad market, saying that "there is no iogicai basis for
distinguishing the Search Ad Market from lhe larger market for Internet advertising. Because a websíte may choose to advertise
via search-based advertising or by posting adverti sernents independentiy of any search, search-based advertising is reasonabiy
interchangeable with other fonns of Internet advertising. The Search Ad Market thus is too narrow to constitute a relevant
market." (KinderStari.corn LLC v. Google, inc.. No. C 06-2057 JF (RS), March 16, 2007. Not reported. 2007 WL 831806
(N.D.CaI))
Before a selier of non-search advertising would be wiliing to adjust the per-click price of its inventory to compensate for any
targeting deficit relative to search ads, the advertising seller would have to take into accounl the opportunity cosI of seiIing its
inventory on a per-click basis: the revenue it would forego from lhe per-impression saie of lhe ads dispiaced.
91
Stephen Shankiand, Google addresses antitrust issue on Yahoo ad deal, CNET NEWS (May 8, 2008), http://news.cnei.com/830 i 10784 3-9939473-7.htrnl.
92
Who 's afraid of Google, THE ECONOMIST (August 30, 2007).
Our review has not found cornpelling evidence that search and non-search advertising do not compete; inoreover, we have offered
affirrnative arguments suggesting that they in fact do.
Benjamin Edelman, Google- Yahoo Ad Deal is Badfor Online Advertising, HARVARD Bu5INE5s SCI-IOOL WORKING KNOWLEDGE,
(August 12, 2008), http://hbswk.hbs.edu/iternl5995.htmi.
°
Benjamin Edeirnan and Michael Schwartz, Optirnal Auction Design in o Multi-unil Environ,nent: The Case ofSponsored Search
Auctions, mimeo (2006). Edelman and Schwarz (2006) show that minimum bids also affect the bids of higher-ranked advertisers
(i.e., advertisers whose bids absent the minimum would exceed that minimum). However, a search-ad selier also faces a constraint
on lhe magnitude of lhe reserve price it seIs, ifan increase in the reserve price would dissuade potential bidders from participating.
Jeremy Buiow and Paul Klernperer (1996) show that lhe expected revenue from an auction with N + 1 bidders and no reserve is aI
leasl as great as lhe revenue from lhe sarne auction with N bidders and lhe revenue-maximizing reserve price. Jeremy Bulow and
Paul Klemperer, Auctions versas Negotiations, (86:1) AMERICAN EcoNoMlc REVIEW 180-194 (1996). (See also Paul Miigrorn,
Putling Auction Theory to Work, CAMERIOGE UNI VERSITY PRESS § 4.4.2 (2004).
' Note, however, that lhe existence of a minimum bid can increase social welfare, notwithstanding an increase in revenue to lhe
search engine. Susan Athey and Glenn Ellison, Pasilion Auctions with Consurner Search § 4 (May 2008).
20
Ft.
Market Definition in On/ine Advertising
opportunity for keyword-specific search-ad prices to reflect differences in valuation between advertisers
seeking particular search terms and advertisers seeking different search terms.`
We note that some researchers have found that better targeting ads increases consumer weifare.98 More
fundamentaliy, the type of differential pricing such keyword-based partitioning aliows is alrnost certainiy
vital to the viabiiity of the online-advertising industry and thus is procompetitive. As in many or most
Internet industries, the dei ivery of online advertising is associated with extremeiy iow marginal costs but
substantial fixed and sunk costs. Aithough in many contexts sunk costs are treated as essentiaiiy
irrelevant—because by definition they have aiready been sunk—sunk costs are not irreievant here or in
many industries characterized by rapid innovation. in rapidiy innovating industries costs are not sunk once
but rather are repeatedly sunk. Thus, if innovation is to continue, there must be an expectation that sunk
costs, as well as fixed and marginal costs, wili be recovered. Baumol and Swanson argue that, even under
highiy competitive conditions, firms in industries with this type ofcost structure "wili beforcedto adopt
discriminatory pricing whenever that is feasibie."99
.
For many advertisers, free "organic search resuits" are a dose substitute for paid search ads. The higher
in the search-resuits iisting a Iink to the advertiser's product appears, the more likely a searcher wiii find
that link. When this occurs, the advertiser does not need to pay for a sponsored position on the searchresuits page.
VI. CONCLUDING REMARKS
Internet advertising has grown very rapidly over the past decade as consumers have shifted their attention
oniine. That growth wili undoubtediy continue for at least another decade, aithough the rate ofgrowth wiii
siow. At the sarne time, it appears iikeiy that there wiii be continuing convergence between search and nonsearch ads. In terms of market definition, we have posed a series of important questions and we have
suggested some possibie answers. In the end, however, much more work needs to be done before we can
determine those answers with confidence.
There is little doubt that offiine advertising has been in decline. Moreover, it is plausibie to expect that
the pricing of oniine advertising is partiaiiy responsible for that decline. To the extent that offline
advertising has faced substantiai competition frorn online advertising, there may no ionger be separate
rnarkets for offiine advertising for media such as newspapers and radio. We are iess confident that offline
advertising constrains the price ofonline advertising; the answer may weil vary depending on the particular
goais of the advertisers and the media that offer the ciosest competition.
Within the sphere of oniine advertising, the issue is whether search and non-search advertising are
sufficientiy competitive so as to properiy be piaced in the sarne reievant market. The relevant question is:
Would enough search advertisers shift advertising volume frorn search to non-search advertisements to
defeat the profitabi!ity of an across-the-board price increase for search ads? We argue that, because
advertisers uitimateiy are purchasing saies, many types of advertising with varied characteristics can
nevertheless compete with each other on price. Further analysis and ernpiricai study are needed before we
have a definitive answer to this question. That study would hopefuliy assist us in understanding the extent
to whieh there is overlap between businesses that buy search ads and those that buy non-search ads.
Overlap itseif does not necessariiy irnply substitutability, but it shouid nevertheiess provide some insight
into the key question regarding switching. Specificaliy, it would aiso tell us whether non-search ad
Goldfarb and Tucker (2007), discussed above, also reported their analysis of a different natural experiment that arises from stateto-state variation in "ambulance-chaser" regulations, which in some states limil altorneys' contingency fees. Goldfarb and Tucker
examined advertising prices paid by iawyers for 174 Google search terras in 195 different locations. They found that lawyers in
areas where contingency fees were limited (and thus the value of a referrai is presumably lower) paid approximately 17% less per
click than iawyers in areas where contingency fees were not Iimited by reguiation.
Esther Gal-Or and Mordechai Gal-Or, Custorni:ed Advertising via a Cornrnon Media Disributor, (24:2) MARKETING ScIENCE
241 253 (2005). Yongmin Chen and Chuan He, PaId Placement: Ads'erfising and Search on lhe Internei, NET INSTITUTE
WORKING PAPER #06-02, (September 2006), http://www.netinst.org/Chen-He.pdf. Asini Ansari and Cari Meia. E-Custornization,
(402)JOURNAL or MARKETING RESEARCFI 131-145 (2003).
William J. Baumol and Daniel O. Swanson, The New Econon:v and Ubiquilous Competitive Price Discrirnination: Ident(,5dng
Defensible Criteria of Markei Power, (70:661) ANTITRUST LAW .JOURNAL (2003). (emphasis in original)
21
Market Dejlnilion in Online Advertising
inventory would be switched from pay-per-impression to pay-per-click in order to serve search advertisers
who need pay-per-click ad in response to an increase in search-ad prices.
22
Markei Definlilon in Online Advertising
References
Alvin J, Silk, Lisa R. Klein, and Ernst R. Berndt, intern1edia Substitulabi/ity and Markel Den,and by Nationa/ Adverlisers, (20:4)
REVIEW OF INDUSTRIAL ORGANIZATION, 323-48 (2002).
Asim Ansari and Cari Meta. E-Cnstornization, (40:2) .IOURNAL OF MARKETtNG RESEARCH 131- 145 (2003).
Avi Goldfarb and Catherine Tucker, Search Engine Advertising: Pricing Ads to Context, NET INSTITUTE WORKING PAPER #07-23,
(September 2007), <tlttp;//ssm.com/abstract= ] 02145 1 >.
Barbara K. Kaye and Nornian MedoffJust a Click Away: Advertising on lhe Internet, MASSACHUSETTS: ALLYN AND BACON, (2001).
Benjamin Edelman, Michael Ostrovsky, and Michael Schwarz, Internet Advertising and lhe Generalized Second-Price Auction:
Sel/ing B,I/ion.s o! Do//ara Worth of Keywords, (97:1) AMERICAN ECONOMtC REVIEW 242-259 (March 2007).
Benjamin Edelman and Michael Schwarz, Optimal Auction Design in a Multi-uni! Environ,nent: The Case of Sponsored Search
Auctions, mirneo (2006).
David S. Evans, The Online Advertising Industry: Econornics, Evolution, and Privacv, (23: 3) JOURNAL OF ECONOMtC PERSPECTIVES,
37-60 (Sumnier 2009).
Esther Gal-Or and Mordecliai Gal-Or, CustonzizedAdvertising via a Common Media Distributor, (24:2) MARKETING SCtENCE 241253 (2005).
Hal Varian, Online AdAuctions, (99:2) AMERICAN ECONoMtc REVIEW, 430-434 (May 2009).
Hal Varian, Position Auctions, (25) INTERNATIONAL JOURNAL OF INDUSTRIAL ORGANIZATION, 1163-1178 (2007).
Jean-Charles Rochet and Jean Tirole, Platj'orm Co,npetilion in Two-Sided Markels, (1:4) JOURNAL OF THE EUROPEAN EcoNoMtc
ASSOCIATION, 990-1029 (June 2003).
Jererny Bulow and Paul Klemperer, Auctions versus Negotialions, (86: 1) AMERICAN ECONOMIC REVIEW 180-194 (1996).
Jianqing Chen, De Liu, and Andrew B. Whinston, Auctioning Kevwords in On/ine Search, (73:4) JOURNAL OF MARKETtNG 125-141
(July 2009).
Leonard Kleinrock , Infor,nation F/ow in Large Communication Neta, RLE QUARTERLY PROGRESS REPORT, (July 1961).
Online
Publishers
Association,
Media
Mix
publishers.org/niedial 1 52Wopamedia_mixstudymar02.pdf'
Studv,
(March
2002),
<http://www.online-
Paul Kleinperer and Jeremy Bulow, When Are Auctions Besi?, NBER WORKING PAPER No. 13268 (2007).
Paul Milgrom, Pulting Auclion Theo,y la Work, CAMBRIDGE UNI VERSITY PRESS. (2004).
Robert S. Pindyck and Daniel L. Rubinfeld, Microecononiics, 7" Edition, PEARSON (2009).
Simon P. Anderson and Jean J. Gabszewicz (forthcoming). The media and advertising: a tale oflwo-sided mar/reI, HANDB0OK OF
CULTURAL ECONOMICS Victor Ginsburgh and David Throsby (eds.)
Susan Athey and Glenn Ellison, Posilion Auctions with Consumer Search (May 2008).
Tim Berners-Lee with Mark Fischetti, Weaving lhe Web, HarperCollins (1999).
Will iam J. Baurnol and Daniel G. Swanson, The New Economv and Ubiquitous Competilive Price Discrimination: ldent(fying
Defensible Criteria af Mar/reI Power, (70 66 1) ANTITRUST LAW JOURNAL (2003).
Yongrnin Chen and Chuan He, Paid Placenzenl: Advertising and Search on lhe Internet, NET INSTITUTE WORKING PAPER #06-02,
(September 2006), <http:/Iwww.netinst.org/Chen-He.pdD-,
23
ADVOGADOS
An Introduction to Marketing
Attribution
!Tí1Eí1tteIuiu [s1 a1iIi1flI Ti1 i iJW1TIiTJ1IL']i]
Selecting the Right Model for
Search, Dispay & Social Advertising
Wt
arketnq ALtribuUoí?
etna lhe, Rg
Te
Attribu0on Niodel
ivtion :o cc; ra v Áttr bo Marke;inc: Sperd
Ch0eaqes ef Áttr hu0c,n
The piture af Âttríution
Enal Pnough15
ÁLoul Marir,
Copyright © 2014 Marin Software Inc. Ali rights reserved.
5 O T TWA ft £
04
04
09
llIi1t1uI9ÍD]I
The battle for revenue is increasingly being fought online. By 2017, advertisers will
spend over $174hn on online advertising in order to generate awareness, attract new
customers, and ultimately drive revenu&.
Despite these massive and growing budgets, many marketers struggle to accurately
allocate credit across their various programs. As this digital landscape becomes
increasingly competitive and complex, marketers rnust find ways to gain visibility
into rnarketing effectiveness, respond quickly to changing market condítions, and
consistently detiver ever-improving financial results.
GLOBAL DIGITAL ADVERTISING MARKET ($BN)
Vaice of the Advertiser
$114
As lhe search auction
irtiensity increases and
CPCs rise, Gross Profil
and ROI gol squeezed
ncieasing lhe importance
ai diversiiying inveslnient
acroIs multlple channels'
Mobile: 23% CAGR
Social: 29% CAGR
Display: 5% CAGR
Search: 13% CAGR
Search
Display
2012
r'
ili [•
0 Social
D Mobile
2011
r*ci';,
alli
3. ,»
CAF, O1-)
The easiest, and therefore often the default method of allocating credit across
advertising clicks is to focus on the ad that was clicked just before the conversion, also
known as "last click attribution." While this may be easy, the last click unfortunately
may not be the only click that deserves credit for a saIe. Marketing attribution refers
to the science or methodology that enables marketers to attribute conversion credit
across every click in the user journey and accurately tailor how that credit is allocated.
The result, if executed properly, is not only the appropriate appiication of credit across
publishers but also budget and bid optimization that is more closely tied to consumer
behavior.
5O! T
n
W AR R
1
Magna Global, Digital Media Forecasts, Decerober 2012
3
play in informing and influencing the customer journey, and subsequently allocating
partial vaRie to different touch points which have influenced a saie or another desfred
outcome.
Voice of the Adveftiser
.
'The pianacle of attril)utina nitelligeticL, wiP Se
rached wher it is
possible to have a siagle
ufliq ue custorner viow
that [adoro ia ali gola ts
of onhne exposure and
IfeUme value
54% OF BLJSINESSES CARRY
OUT SOME SORT OF
ATTRIBUTION
58% BELIEVE THAT
PERFECT ATTRIBUTION IS
IMPOSSIBLE
38% PERFORM
ATTRIBUTION MANUALLY
e 1e krtne: lcd i ecn, 4
't
SELEC1]NG THE RIGHT ATTRIBUTION MODEL
The L.ast Click and F rst-Cíck Mocieis
Marketing attribution started with the last-click model and many advertisers still use
this today. The last-click model assigns credit to the final click in a path to conversion
and does not give credit to any other clicks irvoived in the path. 20% of advertisers
use a last-click modei2 .
100%
90%
80%
70%
t
u 60%
b4E
40%
"3
30%
20%
10%
0%
Akct ri n
% 0f
TWA
a E
First Click
Senond Click
Third Click
2. TagMari, Marketing Agility, Win The Fight Against Wasted Digital Spend
Last Click
4
Tlrst cncç, ine consumer may never nave oiscoverea your orano ano converwu.
+17o 01
agencies and 24% cf brands managing advertising in house use a first click modeP.
100%
90%
o!.",,
ou /o
o,
00
60%
b!3
50%
40%
CS
30%
20%
10%
0%
First CUck
Second Ciick
Third CSck
Last Click
Although iast-and first-ciick modeis remain quite popular, many marketers now believe
they are no ionger sufficient. Critics of iast-ciick attribution rightiy point out that it
falis to give any credit to other ads that may have influenced the saie. Likewise, firstchck modeis ignore he influence of subsequent ads along the path to conversion.
Since ali touch points cari potentiaiiy influence the conversion in some way, credit
shouid be spread across the eritire path to conversiorl. In recent years rnarketers have
begun to evolve toward multi-click attribution modeis, which distribute value across ali
c!icks in the path and enable marketers to make decisions based on the contribution of
each cl ick throughout the path-to-conversion.
AXa ti rI
$0 f T W AR $
3. Econsultancy Marketing Attribution Valuing the Customer Journey
5
:l.
Voice of lhe Advertiser
The clialiengo that
agencies face today IS
ihai clients lave (iiffering
moa suremeni and ai
inhution chailenges and
lhorníoie, require cuslomized solutions. We believe
that lhe conibination ci
data, predictive modeling,
attrbuiiori and real-time
decisioning m biddable
media is a potonlial gamo
charmger for our clients.'
.
ti O;ck Attrbu ori
The Linear Model
The rnost basic of the multi-click attribution models is the linear model, which spreads
credit equally across ali clicks in the path-to-conversion.
Although linear attribution avoids some of the problerns of single-click attribution, it
introduces a different set of chalienges. By spreading credit equaliy across ali clicks,
linear attribution assumes that each step along the way played an equal role in the
conversion and ignores other contributing factors, including latency (eiapsed time to
conversion), awareness and position.
60%
50%
4,
=
bW
O
40h
30%
20%
10%
0%
First Click
Second Click
Third Click
Last Click
Time Decay Modeis
Time decay modeis seek to solve the issue of latency by giving more credit to clicks
that occurred closest to the time of conversion. Time decay models assume that
not ali touch point are equal; however, the contribution of the "introducer" can be
undervalued. In contrast, time decay models can overvaiLie the last click.
60%
50%
•0S
0.5
0.5
ETA
rso'
(SAI
AS
e
20%
10%
((O'
U/0
Aar
FTWARE
-.-Third Click
----,---,-.---.-
Firsi Click
Second Click
Last Click
Copyright © 2014 Marin Software Inc. Ali rights reserved.
[i
Position-based modeis adjust the attribution weight based on the position of the click
in the path-to-conversion, giving more value to the first and Iast click. These models
are designed to avoid over-crediting the first and last click by assigning value across
the click-path. The chailenge with a position-based attribution model is that marketers
often arbitrarily decide how to assign value, rather than using data to inform their
rnodehng.
Voice of the Advertiser
"Switching attribution
modais íoquires ali teams
•withm lhe overaH markelirig furiction tu worh logel hei. In addition b uy-i ri
needs to be secureci troar
fiu rica and mau agem caL
loa as as reportin g wil
change and lhe resufling
nem numbers need lobo
accepted aI ah leveis ia ali
furrLliuns."
60%
50%
40%
30%
20%
10%
0%
First Click
Second Click
Third Click
last Click
Regression and Algorithmic Models
Regression modeling is one of the most advanced forms of marketing attribution,
reiying upon regression analysis of historical performance to aliocate credit. Value is
distributed across multiple chcks, specific to each conversion path. The examples
b&ow and the resulting attríbution are fictitíous and will not be applicable to ali
advertisers; they serve to iliustrate the concept of regression analysis.
In the first example beiow, we have iliustrated a path to conversion where the first click
was a display ad, the second was a generic search, the third was a retargeting ad, and
E
the final click was a branded search.
60%
50%
-o
cc 40%
30%
20%
10%
0%
First Click
Oisplay
Acc l n
SOFTWARE
Second Click - Third Click
heneris Search Retargeting
Last Click
Brand Terrn
Search
Copyright © 2014 Marin Software Inc. Ali rights reserved.
7
A regression model may assign significant value to the first click, recognizing the
significant contribution this click provided in introducing the consurner to the brand
and establishing brand awareness. The model has given the most credit to the second
click a generic search - which played a significant role in the consumer's path-toconversion by leading the consurner to a product page. The third and fourth clicks have
been given less credit as they are leaning on the contribution made from the preceding
clicks. Retargeting relies on the fact that a consumer is already aware of your brand
and the click can be considered to have an element of coincidence. Similarly, the
brand term search can be thought of as a navigational click, deserving less credit than
the second click, but more than the third as it occurred closest to the conversion.
•
The second example considers generic search clicks, a social click and finaily a
retargeting click. In thís instance, the regression model has assigned the rnost credit to
the second click, the generic search. The least credit has been given to the final click,
despite it occurring closest to the time of the conversion, again because of the nature
of retargeting where a consurner is already aware of your brand and products due to
their prior generic searches.
60%
50%
40%
2 30%
20%
CS
10%
0%
First Click
teneric
Search
Secorul Click Ceneric
Search
Third Click Social
Last Click Retargeting
On the surface, regression modeis appear to address the chailenges advertisers face
with rnarketing attribution and fUI the gaps characterized by first click, last click, linear,
time decay, and position modeis. However, there are potential drawbacks that must be
considered before irnplernenting a regression model.
Regression modeis rely on a significant volume of data, and the less data available, the
less accurate the resulting attribution model will be. Making optimization decisions
based on an incomplete ar inaccurate model can result in suboptimal bids that
adversely affect performance. Furthermore, not ali consumers are influenced in the
sarne way by the sarne channels at the sarne time, which rneans that regression modeis
may do a fine job of allocating historical credit but may not necessarily be the best
predictor of future behavior.
5 0 T W A E E
Aigorithmic attribution modeis are also data dependent but learn and adapt
according to the most recent data available. These modeis statisticaHy analyze
and compare paths that resuited in conversions as well as paths that did not result
in conversions, enabling the true contribution of each click and channel to be
accurately modeled.
Voice ai the Advertiser
"If you uflderstfld lhe
Pl YOU CaIi optirrnze lhe
lute re
THE MOTIVATON TO ACCURATELY ATTRBUTE MARKET!NG SPEND
Marketers carry out marketing attribution for three main reasons: to justify
marketing spend, to bufld an understanding of the custorner journey and audience
behavior, and to use this understanding to optimize the media mix. To iliustrate
the impact of attribution, consider the resulting value assigned to clicks when two
different attribution modeis are cornpared side-by-side.
$60
$58
W5
$40
Last COrk
Position case
.52
'- $30
$20
:
Firut COrk
Second COrk
Tflird COrk
Last COrk
In a last-click model (blue), 100% of the conversion value would be assigned to the
last click. The position-based modei (orange) would have distributed the $50 value
of the coriversion equaily across every click. Budget aliocation decisions based on
the iast-click modei would result in reduced spend allocated to the channeis that
drove the first, second, and third clicks. By comparison, budget decisions based
on the position-based modei would result in continued investrnent in the channeis
driving the first, second, and third clicks as welI as the channel that drove the last
click in the path-to-conversion.
Aorn
SOFTWARE
Copyright 02014 Marin Software Inc. Ali rights reserved.
9
be shifted away from poorly perforrning channeis and reinvested in high-performing
channels, The choice of attribution method can have significant impact on budgeting
and decisionmaking, so marketers should choose their model carefuHy.
Channels that drive engagement at the top of the funne!, compared to those that drive
conversions at the bottom of the funnel, require efficient budget aUocation. Armed
with an attribution model, marketers should aim to feed the funnei by assigning
appropriate budget to every channel. Within individual channels, some campaigns
drive awareness and others drive revenue. Marketers should ensure that they are
optimizing not just lhe channel budgets, but also lhe campaign budgets and ad
or keyword bids to ensure good coverage across the entire path-to-conversion.
CHALLENGES OF ATTRIBUTON
Voice of the Advertiser
"Try altering your display
strategy swap your
exisng landing page
tor one better designed
to facihtate a conver
sion, try to sufI dispiay
to dose a saie ralhar
than reiyng ao coe faial
search."
Despite its benefits, only 54% of businesses carry out some form of attributíon4 .
Many advertisers have existing, disparate sources of data and technology that do not
integrate easiiy. This occurs when digital media teams operate in silos and seiect
individual pieces of technology to optimize individual channels without considering
the overail marketing strategy.
For example, assume that an advertiser is running two programs: one display and
one search. When a customer converts after clicking on both a search ad and dispiay
ad, these clicks are sometimes tracked by two different systems. As a result, both
systems track the saie and assocíate it to the click from each respective channel.
Consequently, the advertiser would mistakeniy recognize two conversions, when in fact
there had only been one. Optimízing each channei based on this data would result in
suboptimal performance. Though deduplicating saies across these two channels would
lead to more effective optimization, the data manipulation required is time consuming
and error-prone.
Finding rnarketers wíth the skiiiset to interpret large amounts of data associated with
onlírie media is challenging. Traditionally, rriarketers were hired for their creative
skiils. lncreasingiy, organizations are seeking to híre individuaIs adept in data anaiysis
and interpretation. This two-pronged approach to marketing attribution is critical for
success, but finding the correct biend of analytical thinking and creativity can be
difficuit. As a result, many businesses are unable to tackle attribution effectively.
Internal politics present an incredibie chalienge when implementing marketing
attribution. Resistance and confiicts can arise through fear of change, with each
channel team preferring the channel-centric technologies currently in use. Aiternativeiy,
some teams fear that implementing an attribution model will puil attributabie revenue
away from their channel. Even if ali teams invoived believe that adopting a multichannel marketing attribution model is essential for performance optimization, there
can be issues either justifying or apportioning the cost.
Ma ria
SOFTWARE
4. Econsultancy Marketing Attributíon Valuirig the Customer Journey
THE FUTURE OF
L1l'. Á I/V¼11 .JC151 '...'
LIW L
58% of rnarketers LMI .
1
.
the perfect solution may be unattainable, there are many exciting deveiopments ri
attribution technoiogy.
& AdaptbWty
Fiexibility and adaptability of attribution systems will be key as new channels continue
tu emerge, with new ad formats driving increased engagement. Uitimateiy, businesses
that are capable of accurate measurement, attribution, and decision-making will win
over those Mo adopt a triakmd-error approach.
The future will see the continued emergence of specialist technology designed
specifically to measure, analyze, interpret, and act on consumer data.
Digital marketers need to leam to embrace these technologies and continue overlaying
an understanding of their business to enabie the most effective optimization strategy.
One of the biggest chalienges for marketers will be to optimize programs based on
knowiedge of the future, for example, when switching frorn fuli-price tu seasonal
promotions. Marketers who are dose to their business will understand how campaigns
are likely to react based on past performance, and adjust their optimization strategies
accord i ng ly.
Jo'rg Online & Ofíline
Only 35% of companies incorporate offline touch points into their attribution rnodets6 .
This is an advanced step, but it cari be extremely usefui in driving incremental
performance. Consider the various, common offiine touch points that can be
incorporated: phone calis, tead verification, life-time value, in-store returns, store visits,
and voucher redemption.
If the systems that are capturing ali this data (e.g., cail tracking solutidns, CRM
systems, in-house data warehouses, eiectronic point of saie systems) are abie to
capture a unique identifier that is also present in online media tracking (e.g. order
ID, customer ID, keyword ID and date, ad ID and date) then data can be passed from
the offiine system through the online attribution solution, enabling marketers to gain
valuabie insights that drive informed decisions across the entire marketing mix.
Fi
arui
IOFTWA,E
S. Econsultancy Marketing Attdbuton VaIurig the Customer Journey
5. Econsultancy Mar keting Attribution Valuing the Customer Journey
:[silieB 1 í
The increasingly compiex digital advertising landscape, with a growing number of
channels and touch points, underscores the need for advertisers to build accurate
and effective attribution models. These models, as they become more sophisticated,
go beyond the skill sets of many onhine marketers who have traditionahly been hired
for their creative prowess. To be successful, marketers rnust lean on speciahist
technologies to not only track consumers on their pathto-conversion, but also attribute
revenue intelhigently across the chicks and channels involved in each conversion.
Only hahf of ali businesses today utilize marketing attribution. Those organizations
are rapidly emerging as leaders in revenue acquisition rnanagernent. Advertisers that
ignore or downptay the importance of attribution wihl ultimately be ieft behind as they
will faU to capture the true return on advertising spend and continue optimizing their
programs based on flawed data.
For more informafion:
i
\
í[email protected]
insights. marinsoftwa re. com
jMarinSoftware
Akarín
S O F T W AR E
Copyright © 2014 Marin Software 1 nc. Ali rig hts reserved.
12
(1
With more bian $6 billion in annualized advertising spend, Marin Software (NYSE: MIRIM is the
eading digital ad management platform in the world. Offering an integrated platform for search,
display, social, and mobile advertising, Marin helps the world's hest brands and agencies simplify
their advertising workflow while dramatically increasing ad performance. Powering advertising
campaïgns in more than 160 countries, Marin's technology transforms data into insights and
complexity into opportunity for hundreds of global advertisers and agencies. For more information
.
about Marin's solutions, please visit: http://www.marinsoftware.com/solutions/overview.
Unted States
San Francisco
123 Misson Street
j
EMEA
APAC
United Kingdom
Singapore
lst Floor, Orion House,
One Raffles Ouay
25ffi Floor
5 Upper St Martin's Lana,
North Tower, Levei 25
San Francisco. CA 94105
London, WC211 9EA
048583
Tal: +1 (415) 399-2580
Tel: +44 (0)845 262 0404
Tel: +65 6622 5888
New York
France
Japan
215 Park Avenue South
Actualis Levei 2
3rd Floor, Sanno Park Tower
Suite 1801
21 and 23 Boulevard Haussrnann
2-11-1 Nagata-cho
New York, NY 10003
Paris 75009
Chiyoda-ku
TeL +1(646) 490-2427
Tel: +33 1 80 95 68 60
Tokyo 100-6162
Tel: +81 3 6205 3179
Chicago
Germany
140S. Dearborn Street
Suite 300-A
Am Kaiserkar 1
Australia
Levei 33
20457 Harnburg
Australia Square
Chicago, IL 60603
Tal: +49 40 80 80 74-522
TaL +1(312) 267-2083
264 George Street
Sydney NSW 2000
Tel: +61 (0)2 9258 1820
Austin
800 Brazos Street
Suite 320
Austin, TX 78701
Tel: +1(512) 287-7478
11
II
FIS••--Q-(245.
I
•
1(oWili
7014
4
Veda Beatriz
tfr4
. S. Singer Afonso
l'PGG
DAS-1 f, 2.2 CGAAt
]
_
-O
;..,
-C
f•J
ai
c)
-c
.--,
..2;
o
.<1
C.).
.,
cwip
4
i
0.1
C•1
O
r:
'4!1
•,-,
~
04
g
C4
-aci
cd
-c7'd
©
c
ai
g
oo
cv
ooi
o
kr)
kr).
cnI
,--i
cn1
o
kr)
kr)
r-cn
@
@
cd
cd
cd
Z
Z
Z
oo
cv
oo
cv
oo
.
i--1
cn
efi
il9
••■
c>
EO
ccd>
,sc>z
cl)
i
i--1
en
en
1
,--i
O
O
0
-c u
O
C
O
u
en
1--
en
1
,--i
cd
o
o
o
ç
C.-)
O
c>
r2
O
-41-d
cd
çà
ai
-c
u)
›;.0
O
ai
ai
c
Z
o
ã
e
XI
c>
i
o
ã
c>
A
(.5
ai
O
çà
O
o
c>
Yedda Beatriz G. A. D. C. S. Sinqer Afonso
De:
Enviado em:
Para:
Cc:
Assunto:
Marcelo Nunes de Oliveira
quarta-feira, 6 de agosto de 2014 16:35
, c3
Ricardo Motta 1 GCBA
Yedda Beatriz G. A. D. C. S. Singer Afonso; Leonor Cordovil 1 GCBA; Pedro
Yukimitsu Ribeiro Tokuzumi
RES: Ofício 2316/2014ICADE/SG/GAB - Google (PA 08012.010483/2011-941 e
08700.005694/2013-19)
..__-
Ricardo,
Conforme combinado, defiro a dilação do prazo nos termos propostos.
Att.,
Sarcelo Nunes de Oliveira
Coordenador-Geral de Análise Antitruste
Superintendência-Geral - CADE
Tel.: 61-3221-8515
De: Ricardo Motta 1 GCBA {mailto:rcmgcba. com. br]
Enviada em: quarta-feira, 6 de agosto de 2014 15:43
Para: Marcelo Nunes de Oliveira
Cc: Yedda Beatriz G. A. D. C. S. Singer Afonso; Leonor Cordovil 1 GCBA; Pedro Yukimitsu Ribeiro Tokuzumi
Assunto: Ofício 2316/2014/CADE! SG/GAB - Google (PA 08012.010483/2011-94 e N° 08700.005694/2013-19)
Ao Sr. Marcelo Nunes de Oliveira,
Coordenador-Geral de Análise Antitruste 1
O GOOGLE BRASIL INTERNET LTDA. (Google), por meio de seu advogado devidamente qualificado, vem,
peitosamente, requer a concessão de prazo adicional de 15 dias para apresentar resposta ao item 6 do Ofício
6/201 4/CADE/SG/GAB referente aos processos administrativos N° 08012.010483/2011-94 e N°
08700.005694/2013-19, tendo em vista a dificuldade encontrada em coletar e organizar os dados de modo solicitado
por essa Superintendência.
Atenciosamente,
Ricardo Casanova Motta Grinberg e Cordovil Advogados
Alameda Jaú, n° 1742, 6° andar— CEP: 01420-904 - São Paulo - SP
Tel. Dir.: +55 113371-5054 j Telefone: + 55 113371-5050 1 Fax: +55 113371-5059
www.gcba.com.br
Esta mensagem (incluindo qualquer anexo) Contém informações confidenciais para fins específicos e é protegida por lei. Se VOCê não é o destinatário desta mensagem,
você deve deletá-la. Qualquer divulgação, cópia ou distribuição desta mensagem é estritamente proibida.
This message (including any attachments) contains confidential information intended for a specific individual and purpose and is protccted by Law. If you are not the
recipient, you should delete tbis message. Any disclosure, copying or distribution of this message is strictly prohibited.
1
SI
Cnse1bo
O E /\
SISCOP
de Delesa
Solicitação n°: 21735
Autorização de Copia do Processo n°: 08012.010483/2011-94
Nome do Solicitante: Gabriela Miranda Naves
Nome da Empresa:
Número do documento: 00554398109
Telefone de contato: (61) 3321-2105
Tipo do Processo: Procedimento Administrativo
C;Ç)
Número de foãias\,
Valor:
Partes: E-Coxrnrce Media Group Informação e Tecnologia
Ltda., E-Comrnarce Media Group Informação e
Partes para cópia:
Peço cópia dos CDs anexos na Petição n° 08700.005821/2014-60 da Google emresp. ao o n°2316/2014 (Fis
1922/2034).
cO
O Usuário não tem procuração
De ordem, indefiro.
49
Data: / /______
z G.
1
Assinatura! Carimbo (Obrigatório)
0flO
EPPGG
VDAS..i02.2cGM
Recebimento do solicitante
Solicitação atendida emO6/ 0I'I ('
Nome por extenso
Assinatura
__
JSJ
Yedda
\5
,
>O
a)
0
a.)
o
çn
-coa)
--(
-
0
Cd
a.)
E
O
0
E
cú
01
c_l
.0
cn
Cd
O
()
(
a)
@)
o
@)I
a)
z
00
00
00
1
00
00
00
1
1
LO
O
LO
O
LO
O
LO
1
Çn
cfl
1
1
C)
CO
1
C)
CO
a)
o
1
Z3
cri
-o
-o
a)
o
413
..0
O
a
LO
/)
c
c
bi
C)
CO
-
1
-
O
tD
O
1
L1 '—
O
O
LO
1
N
1
O
-
-
E-U
1
c
. a)
a.)
)
z40,
O
o
o
-a)
as
.
.1-Ci)
o
o
o
—J
o
•0
o
O
C
— .4 ------'
1o
or.
L)
-
k-s.
I
G
c6n
oe /'f';'.
Yedda BeamÇS
D
TT TTCTrPTC.T1ÂCb 'pTTT-TflP
MAR
T
fl NUNES
PGG
A!1
Fis
flF flT TVPTPÂ CflCbRF1FT\TÂF
GERAL DE ANÁLISE ANTITRUSTE DA SUPERINTENDÊNCIA GERAL DO
CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA - CADE
CADE/MJ
Protocolo - Geral
fl
08700.006835/2014-00
Processo Administrativo n. 08700.010483/2011-94.
GOOGLE BRASIL INTERNET LTDA. ("Representado"), já qualificado nos
autos do presente processo administrativo, por seus advogados infra-assinados, vem,
respeitosamente, requer a juntada de cópia de apresentação de slides realizada em reunião havida
no dia 28 de maio de 2014 nessa Superintendência-Geral do CADEtAproveita o ensejo também
para trazer, abaixo, breve sumário dos principais pontos discutidos em reunião.
1
A. Evolução das Ferramentas de Busca - Breve Contexto Histórico
1.
A principal missão das ferramentas de busca é conectar os usuários com as
informações que estão sendo procuradas, da forma mais rápida possível (slide 5). A despeito de
as ferramentas de busca terem iniciado suas atividades de organização das informações na
internet por meio do simples ranqueamento e fornecimento de links azuis, sem dúvida esse não o
único tipo de informação disponível e pesquisada pelos usuários (slide 6).
2.
Quando os usuários fazem pesquisas em ferramentas de busca, eles não estão
necessariamente interessados em encontrar linIes aiis como resposta para essas buscas. Os
usuários, em regra, esperam respostas diretas, logo na página de resultados, a fim de encurtar o
tempo de busca e evitar que a mesma pesquisa seja repetida em outras páginas da internet para
que se possa encontrar uma informação relevante. É consenso entre as ferramentas de busca não apenas o Google, mas também o Bing e o Yahoo! - que quanto mais esclarecedores e diretos
os resultados apresentados aos usuários, melhor será a experiência de busca. Ao fazer uma busca
na internet, os usuários, de uma maneira geral, podem estar procurando por diferentes tipos de
informações, tais como imagens, mapas, vídeos, clima, dentre outras. Ao invés de realizarem suas
buscas em diferentes páginas da internet - urna para cada tipo de conteúdo - os usuários podem
fazer pesquisas em uma única página e receber os resultados mais relevantes, independente do
tipo de conteúdo procurado.
3.
Resultados temáticos, tais como vídeos, imagens, mapas, compras, bem como
respostas diretas como o clima, horários de voos, ou horário de filmes são todos modos pelos
quais as ferramentas de busca fornecem aos usuários resultados mais úteis e ricos (slide 6). Tanto
é verdade que as ferramentas de busca acreditam que respostas mais diretas são em regra
melhores para os usuários, que outros concorrentes começaram a exibir resultados temáticos
sobre produtos muito antes de o Google Shopping ser lançado no Brasil, no final de 2011. Como
mostra o screenshot (slide 7) abaixo, o Yahoo!, por exemplo, já possuía resultados de compras em
2001 no Brasil.
2
Ba
AI d,Aoe,e,
Ezclu,wu x,e
Su900l
te000flIS0000
Crie e grrerscíe seu f&om de cfiseusçãea por e-ermO no XohcoiQnwn, uru os arungor c esperais
Croepete 1 lan LjatOg SiorOOrjotO 11.°ognis e1.0le 600044e4 FJfl00*00 iO1LOA SmsmOo1eueun Iaisaattteo
ounidud., fiot4otoo't fnoOuL Çleikm otcr1es Lm~— funoolawourVOati Leseo GeeSotrea Uuopatt
5504 moia».
Zs4oxroo Z!ïob.eo
Agendo lrZrdnr Cessgeniee ZusuO ZoOonxeu Zuos L
Lojas
lYahoo! Shonctroo - s, ro
Deucoques Livro, em destaque
Lismoaabei Ms
Aruenretras
fru040dLbotto
CD..'adurasoCrlrradsota fotO!
moePoiWca
___
Aoimoror
dnEVOpurRS15O0
uenna gogeoisemo
Informática e interneit
SV.bSdbNacôcios e Economia
Ciênciais HUM nas
Eatneoeif4Oeia ffelatSa
82 Gnutooe» itOgeaflmrittoo»fiutti±goto
Edticeco eFormpco
am3tgto055 focel
NotIrias
momo
Agnuer Zuniras
Entratenímonto
glQu8j
Oito jogOra Çm o4Ozg. olrarer -. fotge5Otdadrr5tnogeege
4.
tlotictau em destaque
Cosmo do' raromumnleugeru
lmouorJaotmounto.t
UotaoZufeumOo'
oaeunuo oenLoenoro
uste*n oMcmldrsenun
Por destoo do Vafool
MMw! Cmdlrusd 0l»41»e e
Cuuh.çaaesar001tOt da 00050
OnÇiola- co. tua trome-page
orlua nOi ,p.eo' vumrar
Shrtdort oullaaes de catOte
pan; gota ladas as
8OttaOAPto pet5açe 1.
Ltao,mrrotõet e fole
A Microsoft, por sua vez, passou a exibir resultados de compras em 2007 (slide 8).
/Jeh
3O 111
Suo alor !iSlltS. ~,
É! t400 tieto. t
Miçroott lela
35 coe o'aasu san
Zu1!IDOVO1kadS G
Ttp RaIod. ido Ltwoist Oses llXl Milltos MP3s, Munas, Gamas TV Showe
Zune Fienlews
fone Software
Zuno Accessoites
ts(aQL4!
s,jtOar
8sf Uc'esoft Zune id Olke Depot f Paceun 14 Dal Frue Zurro PaoV
S
rur 'uotr5O unte
Dearriload J mOmi Freo Motc 140 DOwOload Fnoo Dosottoad Zuna 1Mw
Zoo Vidas
2 tosu Mark.tploCe
luva Saies
Lour fitem.
Microsoft Zuno digital piayer
Resultados
de Shopping
da Microsoft
$t85
fAZ 41 8qy
MSN $hoppm
d,94.1 modu
flrl oddo o IwrV 140 çu e
selated 1oll-Iesth orooplo rroc10, playlsto, polutoS SI 00'
wmagrown uostto recll from 1j*1p.
sf010
Oesr toutewe ato qiance 11631
140* 04oluble te fla' S
oaun Shorn, IV Podol & Mor
OlMo 51*0 1r0e »hlppn9 no 00*0*0
$24 1 w o pl s e 01010
coO, cxu
Sraaur,
XAno
1l Saiwsre Topa Gol Lrrdn O Mor.
-coo hppu9 0 Sadri froer lISO
Zune noi 1 Horr,
W 1cpme to 1110 £xlal
0111 uol roto for Uru Mtcrozofl Zune
o'0s 0mw SOl
Zune flO0l
Shop oS lhe Gol MP3 Pl»o»o fiara
Soe lhe Vlpe ood Orrarol
000* Coo f'øt'50 L5hnsoto, o
W'0001 VISO eStIro» tromOol OlMO
a000
*050
5010*
lolIot UlMor
&ÇS °
00005 Cl000
0005151
Zune MP3 Plavr 1 Zune 2 NOws
MP Plapuo (000 Tt. O Miro
lllal*rnd ordem nor CO ehp fruo
Hoje, são inúmeros os tipos de resultados temáticos exibidos pelas ferramentas de
5.
busca no Brasil. Todos esses tipos de resultados são apresentados pelas ferramentas de busca
como forma de fornecer respostas melhores, mais diretas e úteis para os usuários (slide 9).
B. Como Funciona o Google Shopping
6.
•
No www.google.com.br, o Google Shopping é exibido como um grupo de
anúncios de produtos (Product Listing Ads - PLAs) que aparecem nos espaços das páginas de
resultados (i.e., parte superior e lado direito da página de resultados de busca) em que o Google
tradicionalmente exibe anúncios. Os PLAs são anúncios que exibem a imagem de um
determinado produto, o preço pelo qual determinado varejista está vendendo o produto e o
nome do varejista. Os PLAs são gerados por meio defeeds de dados dos varejistas fornecidos ao
Google, via Google Merchant Center (slide 13 a 15).
7.
Por ser uma forma de anúncio, os produtos que são exibidos nos PLAs são
devidamente identificados como "patrocinados" e/ou "anúncios", e não há qualquer interação
deles com os resultados de busca orgânica do Google. A precificação dos PLAs é feita por meio
do custo por dique (CPC), definido por meio de leilões, nos quais os varejistas ofertam seus
•
lances. Entretanto, a ordem de ranqueamento dos produtos na página do Google leva em
consideração não apenas o valor do lance de cada varejista, como também a sua relevância e
qualidade.
8.
Todos os varejistas (ie. websites que vendam produtos diretamente aos usuários)
podem comprar PLAs. Os varejistas que compram PLAs não são obrigados a enviar todo o seu
catálogo de produtos para o Google (embora eles possam). Os varejistas precisam apenas enviar
ao Google as informações referentes aos produtos nos quais desejam anunciar no formato PLAs,
por meio do upload de feed no Google Merchant Center (slide 15). Por exemplo, para aqueles
produtos que o Buscapé comercializa diretamente ao usuário no seu próprio site, o Google
permite a compra de PLAs e o Buscapé deverá apenas fornecer feed dos produtos (informações
relevantes para o anúncio) que deseja realmente anunciar nos PLAs (slide 27).
4
9.
As informações solicitadas pelo Google - e fornecidas pelos varejistas por meio dos
feeds de produtos - são padrão e estritamente relacionadas aos produtos que se pretende anunciar
(slide 14). Esses dados são importantes para que o Google possa construir, de um modo
automatizado, o anúncio no formato de PLAs. A titulo exemplificativo, o Google solicita
informações como (i) identificação do produto; (ii) título; (iii) descrição; (iv) link; (v) condição do
produto; (vi) preço; (vii) disponibilidade; (viii) imposto e frete. É importante notar que o Google
não solicita informações relacionadas a produtos que os varejistas não desejem anunciar nos
.
PLAs.
10.
Conforme destacado acima, sob a perspectiva do usuário, os PLAs são
devidamente rotulados como resultados "patrocinados" do Google (slide 17). Ainda, o Google
separa os PLAs dos resultados de busca orgânica exibindo-os nos espaços da página de resultados
de busca (parte superior e lado direito da página de resultados de busca) em que o Google
historicamente exibe seus anúncios, conforme é demonstrado nos slides 17 e 18. Vide abaixo:
Resultados no Google Shopping
P1rnunado
açuv'
Gom base em sua consulta de
tentando encontrar um produto Ao (Ixv
à
i'
resutados dc a'guns forneccdoe co pcJm a1tr a
d
solicitação O Goope poda s'r rem
.
k
fornecedores.
1V LED 3D 46"
11.
Sarnsung..
TV LED 3D 46"
Samsung
1V 3D Stim LED
46 Fufl HO
R$2.298,00
Walmart
R$1.999,00
Sarawacorn br
R$2.499,89
Extra.combr
Os PLAs, exibidos com preços e imagens, direcionam os usuários que dicam nos
anúncios diretamente ao site dos varejistas onde eles poderão comprar os produtos anunciados, e
não para a página do Google Shopping. O único link que direciona o usuário para a página do
Google Shopping apresenta apenas a expressão "Resultados no Google Shopping" (vide acima),
não havendo qualquer imagem ou indicativo de preços relacionados a esse link especifico.
5
/E PD
ÍÍ'ííÍ
'
O Google apenas exibe resultados de varejistas quando relevantes para a pergunta
12.
do usuário na caixa de pesquisa do Google. Se nenhum resultado do banco de dados do Google
(alimentado pelos feeds enviados pelos varejistas) é relevante para uma determinada pesquisa,
nesse caso nenhuma oferta de produtos será exibida, mesmo que, por exemplo, os algorítmos de
busca do Google interpretem que um site de comparação de preços possa ser relevante para essa
mesma pergunta (slide 19). Isso ocorre porque o Google não exibe PLAs como respostas para
buscas sobre sites de comparação de preços, ou mesmo como resposta para pesquisas
relacionadas especificamente aos sites Buscapé ou Bondfaro. Os resultados de busca do Google
apenas exibirão PLAs se esses anúncios específicos forem relevantes e relacionados a algum
produto específico ou à intenção do usuário em adquirir esse produto.
C. Transição para o Modelo Pago
13.
O Universal Search, que incorpora os resultados de busca temática dentro do
Google Busca, foi lançado em 2007 nos Estados Unidos. Nessa época, o Froogle, ferramenta
originalmente lançada nos EUA em 2002, teve seu nome alterado para Product Search. O
Product Search - também conhecido como Google Shopping - era gratuito para os varejistas e,
por conta disso, era exibido dentro dos resultados de busca orgânica do Google. Em 2009, os
o
PLAs foram lançados nos EUA como uma versão beta.
14.
Em 2011, o Google Shopping foi lançado no Brasil como resultado temático da
busca orgânica, sendo, portanto, exibido, assim como o Product Search nos EUA, dentro dos
resultados de busca orgânica do Google. Pelas razões abaixo expostas, em 2012, o Google
migrou o Product Search nos EUA para um modelo de formato de anúncios e os PLAs
substituíram parte do resultado da busca orgânica (lançados em 2009).
15.
No Brasil, a mesma transição do Google Shopping para os PLAs começou a ser
efetuada em fevereiro de 2013, tendo sido concluída em junho do mesmo ano. As razões para a
transição do modelo gratuito para o modelo comercial podem ser assim resumidas: (i)
necessidade de melhor competir com concorrentes como a Amazon, que ganharam uma fatia
6
Ffs
maior de buscas por produtos em comparação com as ferramentas de busca; (ii) necessidade de
exibição de uma experiência unificada de compra ao invés de uma experiência de busca duplicada
do Product Search e dos PLAs; (iii) incremento dos benefícios para os usuários com informações
mais precisas; e (iv) necessidade de incrementar os benefícios aos varejistas com informações
mais precisas e simplificadas, maior controle e previsibilidade na exibição de anúncios.
1) As Inovações da Amazon levaram muito mais usuários a iniciarem suas buscas com
o
intenção de compra na própria Amazon, ao invés dos sites de busca (slide 25)
16.
A Amazon se tornou um dos maiores players de compras nos EUA ao longo dos
últimos anos, oferecendo aos usuários variedade, preços baixos, compras por meio de um dique
e entrega rápida. A Amazon aprimorou o conceito de marketplace no mundo ao oferecer para os
usuários uma gama significativa de produtos próprios e de outros varejistas. Em razão das
inovações introduzidas pela Amazon, cada vez mais os usuários começam suas compras na
própria Amazon, em detrimento dos sites de busca. Em resposta a esta pressão competitiva, o
Google decidiu fazer a transição dos resultados orgânicos de produtos para PLAs pagos.
17.
Um grande exemplo dessa facilidade dos usuários em encontrar e comprar os
produtos que estão pesquisando é o botão "Buy now with 1-cick", disponibilizado pela Amazon
em seu site de compras (vide abaixo). Por meio dessa ferramenta os usuários podem, com um
único dique, comprar facilmente qualquer produto que estejam procurando no site da Amazon.
Frise-se, novamente, isso tudo por meio de um único dick. Por conta da facilidade e comodidade
de compra e da experiência positiva proporcionada para seus usuários, a Amazon experimentou
crescimento significativo no mercado (slide 23).
7
Fis
amazoncom.
18.
Essa tendência de crescimento, por exemplo, pode ser observada no slide 24, a
partir do gráfico que analisa o crescimento do 'início de compra' na Amazon (ou seja, onde os
usuários começam a fazer compras on-line) em comparação com todas as ferramentas de Busca
conjuntamente, abaixo (slide 24 e 25):
Au,azon
Sarcb EnIns
30%
fl
2009
2012
2009
2011
Fonte: "Why Amazon Matters Now More Than Ever," Sucharita Mulpuru and Brian K. Waiker, Forrester Research,
July 26, 2012, Figure 2.
19.
Conforme se observa do gráfico acima, o início de compras na Amazon cresceu
12% de 2009 a 2011. Em 2009, 18% dos usuários iniciavam suas compras a partir do site da
Fis
5h
Amazon. Em 2011, 30% dos usuários passaram a iniciar suas compras na Amazon. Esse
crescimento da Amazon foi acompanhado pela queda do início de compra nos sites de busca
(Google, Microsoft, Yahoo! e outras, conjuntamente), que decresceram de 24% das buscas de
início de compra, em 2009, para 13%, em 2011. Devido ao aumento da concorrência a partir de
sites varejistas como a Amazon, o Google entendeu que tinha de melhorar a qualidade de seus
resultados para buscas com intenção de compras e decidiu fazer a transição para PLAs.
20.
No Brasil, há diferentes tipos de sites que competem por buscas com intenção de
compra. Não há competição apenas entre sites de buscas e sites de comparação de preço.
Varejistas como Mercado Livre, Americanas.com, WalMart e Submarino, bem como Bondfaro,
Buscapé e outros sites competem por usuários com intenção de compra (slide 26). Assim como
esses sites, o Google, como um agente organizador de informações para os usuários também
deseja atrair usuários interessados em compras. Importante frisar que o Google vende PLAs a
qualquer site, mesmo aqueles que competem com o Google, desde que eles comercializem
diretamente aos usuários em seus próprios sites os produtos que desejam anunciar.
2) No passado. o Product Universal (nos resultados de busca orgânica) e os PLAs
(resultados patrocinados) eram exibidos de forma duplicada na página de resultado
o
de busca nos EUA;
21.
Nos EUA, a exibição de respostas de shopping na página de resultados do Google
remonta à implementação do Universal Search, em 2007. Paralelamente, em 2009, o Google
lançou, também nos EUA, os anúncios em formato de PLAs. Com o lançamento dos PLAs, os
resultados de produtos começaram a aparecer em dois lugares separados na mesma página de
resultados, criando uma duplicação, conforme demonstrada no slide 29 (vide abaixo):
9
CORDOVIL
ii.
ADVOGADOS
$i.wp M'ii Orirp çlor,rtap tOrpe -
• Vir,raat Srwn4. lmoor, Mapa Vera
o-
Search
Ar
—
Alta poiais! lo (50*
Qanofoebi
V,dsr,e
Ipaoap4
pa-kbiTe**$
890.90
Meijer
Neare
Qa!OpOp
p$pnp
938-00
Walanapt
s,005iap
San M.taa. CA 94484
Clianga oPoria.
Any tIres
Peot troar
tolp( 24
P0012 danla
Real—k
Real i"-t
Pesi
p
AlI *0.0115
9,10*50 çoarchn
More tOa.O9 boIs
Mreirltren TrI
correra Hikø
$Fnai
824.98
Areerap.sonn
QaRigltt
SUV Iepi
8290.09
CarapbrgWo
Ihagq opa?
ipautpiilpat.
.
.!*lal *heoe Ode?
EureK TontA for CartIOrnO
— baakpsakt.*rt.prooaorrslEorakeT.r*t.
tora StrÇOrIr.rJ and troa tr000eaOrp (ia
M 0.40,5 (irAn 115,5 part tpp&
b~piaap
r.Qakin
1.
l<nWt
Retatord 800rOtios for (artE
Claras CCI CCibniat 0pa!a01 &ns!aa Canpapn1ix
Ura.,dc Çp4pmap Eareb8 IS1tC f4piIb.Eaç2 Qobimbio
Cotpman -Tent TonI 1 Fanctv TortO 1 Camnlng Tent Backnactitnc
—t— P~11 1000
Recaira 1 -7 pf 7- Loolrinp (ore *5.0110 (55*. bo rtrÁirrq teci ar narnp.rsp toat' Vai
COVO1ot000r to (ir.! ~490 Caiarrian ta.,t te.
ta.t.ot Testa Elite ToOto Saakpaok.np TarAS 7.0* AraPdaa
-
1ingJe$tIli3.for.ient
Q~ tratara 14. te, ra,- ForAS. panoon Troo Rase, Terrt
5178. 14 alores
raFtâdStN Te 0.005
Rapar ton)
www.rotoar,4'T.nts
.-ry. mtod
Our Tens Salnctiorn trriroenaive
Orna $toppinp co, 0.40,0 Doar $81!
• $ar*rraapl.00fs ftprn 0001
Tente for Soto
onera otrornyte550 000,1
Anreritar Moda anal ir,
Pala tanta tramo E r005lonTenta
Wtbimai Tenta
— sOdrnert.00rrrdOatdpOrri.lr,lrrp
o,tsd'
SlwiraVerltOy r,1Teot.ForVr.rir
C.nrnpinlr troado ai Wolnra,*
17
7500004
O10 "~Prai Tloca Opor,, A5riniy..n...,
$130- 66 alores
-
Ien* WitdpeIa. Uir2Jme eflCyCtqS30lIt5
kadl.orlwltd1Taot
A teci Iseaheltar ppire,at,ng aí salsada r/ (abria 5* oilroro,alarii 4r05.0 Ova. alt.Chucr
tua (carne aspaS. o, altachori la o sopportlrig 1000 Opala omaliar tenta..
22.
Com o tempo, o Google percebeu que poderia melhorar a experiência do usuário
ao possuir um único local na página destinado a resultados de shopping, ao invés de uma listagem
orgânica e outra listagem paga, exibidas em seções separadas da página de busca. Esse foi um dos
primeiros passos para a transição para um produto que oferecesse uma experiência de compras
única para os usuários.
3) Os usuários são beneficiados com dados mais precisos dos varejistas;
23.
A partir da experiência prática, o Google percebeu que quando os diques eram
gratuitos (i.e, os varejistas não pagavam para terem os seus produtos exibidos no Product Search),
alguns varejistas deixavam de atualizar dados com a frequência e regularidade necessária. A
transição para um modelo pago proporcionou incentivos para que os varejistas mantenham seus
feeds de dados atualizados. Quando os varejistas não mantém seus feeds de dados atualizados, eles
acabam pagando por diques que não levarão os usuários a efetivar suas compras em seus sites, o
que ocorre, por exemplo, quando não há estoque do produto, ou o preço de venda não
corresponde àquele anunciado.
10
A alteração, portanto, do modelo gratuito para o modelo pago visou criar os
24.
incentivos necessários para que os varejistas fornecessem dados ainda mais precisos,
beneficiando, com isso, os usuários que recebem melhores resultados de produtos em respostas a
suas pesquisas no Google (slide 32).
4) Os varejistas queriam um produto com maior simplicidade, no qual eles pudessem
ter maior controle e previsibilidade sobre a exibição de anúncios;
25.
Por fim, a transição para o modelo de PLAs proporcionou aos varejistas maior
controle sobre os produtos que eram exibidos e previsibilidade quanto ao tráfego recebido (slides
34c35).
26.
Quando o Google Shopping era gratuito, os varejistas faziam o upload do seu
inventário de produtos, mas não tinham como indicar quais produtos eram mais
importantes /relevantes para serem exibidos. Em outras palavras, os varejistas não possuíam
nenhum modo de indicar ao Google quais produtos eles desejavam efetivamente promover (slide
34).
1
27.
A transição para os PLAs - com seu modelo pago - permitiu os varejistas, por
meio da realização de leilão individual para cada produto, alcançar (i) um maior controle sobre os
produtos que desejam exibir com mais destaque; (ii) a habilidade de gerar tráfego adicional para
seus sites; e (iii) maior controle e previsibilidade sobre o tipo de tráfego recebido. Nesse sentido,
os lances do leilão são um importante indicativo que os varejistas podem empregar para colocar
em destaque aqueles produtos que eles pretendem promover, sinalizando ainda aqueles produtos
que são lançamentos /novos. A partir dos lances ofertados no leilão, o Google pode levar em
conta essa informação na hora de determinar os produtos relevantes para uma determinada
pesquisa de um usuário (slide 35).
11
Fs
28.
A transição de um produto gratuito para um produto pago teve uma dupla
finalidade, beneficiando tanto usuários quanto varejistas.
29.
Para os usuários, os benefícios incluem a eliminação de resultados duplicados por
meio da exibição de um único produto com o avanço da qualidade e precisão dos dados.
30.
o
Para os varejistas, a transição do modelo gratuito para o modelo pago trouxe
benefícios ao se proporcionar maior controle sobre os produtos específicos a serem exibidos, o
que, por sua vez, resulta na maior previsibilidade do tipo de tráfego recebido. Desse modo, os
varejistas ganharam a habilidade de se diferenciarem por meio da identificação daqueles produtos
que eles pretendem destacar, levando à geração de um tráfego de melhor qualidade. Houve,
portanto, benefícios mútuos para ambos os lados da plataforma do Google decorrentes da
transição para os PLAs (conforme apresenta o slide 35).
D. Por que os PLAs são projetados para Varejistas que vendem produtos diretamente
de seus próprios sites
o
31.
Desde o seu lançamento nos EUA, no ano de 2009, os PLAs foram projetados para
varejistas que vendem produtos diretamente aos usuários em seus próprios sites. O Google
desenvolveu os PLAs por uma razão simples: os usuários querem respostas diretas, rápidas e
convenientes para suas buscas, ao invés de terem que fazer múltiplas pesquisas ou visitar diversos
sites (um a um) para encontrar as informações que estão pesquisando. Os PLAs foram
desenvolvidos para endereçar as expectativas dos usuários. O PLA é um tipo de anúncio que
exibe um produto específico, a um preço específico, e com a imagem desse produto específico.
Quando os usuários veem um preço ou uma imagem de um produto específico nos PLAs, eles
esperam, ao cicarem na imagem, ser direcionados para uma página na qual eles possam comprar
aquele exato produto anunciado. Se o vendedor do anúncio exibido nos PLAs não fosse um
varejista que vende o produto em seu próprio site, então os PLAs não serviriam para o propósito
para o qual eles foram criados.
12
32.
Os PLAs simplesmente apresentam aos usuários uma outra opção para
encontrarem e comprarem produtos. Quando a intenção do usuário é procurar um determinado
produto, é consenso que as ferramentas de busca devem, além de exibir links azuis, oferecer
respostas diretas para a pergunta do usuário. Direcionar o usuário para outro site para que ele
tenha que iniciar novamente sua busca é inconveniente e um verdadeiro desperdício de tempo
para o usuário - exatamente o oposto de uma resposta direta, rápida e conveniente.
33.
Além disso, o Google busca oferecer a melhor experiência de busca por produtos
para seus usuários e, para que isso possa acontecer, é essencial ter um contato direto com os
varejistas. Esse contato direto, sem intermediários, permite ao Google, por exemplo, suspender
um varejista cujos preços anunciados não coincidam com os preços reais ofertados no seu site.
Os usuários têm uma experiência de busca pobre e ruim quando o preço ou produto anunciado
no PLA não corresponde ao preço ou produto no site para o qual ele é direcionado quando dica
no anúncio. Por essa razão, o contato direto do Google com o varejista permite ao Google tomar
medidas apropriadas para que os usuários não sejam enganados ou prejudicados.
34.
Conforme demonstram os sreenshots apresentados no slide 38, outros sites de busca,
como o Yahoo! e o Bing, também exibem anúncios de produtos cuja participação é limitada a
varejistas. Enquanto o Bing e o Yahoo! Poderiam, inclusive, obter receitas maiores se vendessem
PLAs a não-varejistas, eles também presumivelmente concluem que é melhor para seus usuários
direcioná-los a um site no qual eles poderão comprar um produto ao dicarem em um anúncio de
produto.
13
YAHOO!
—
S*ofr
P
bwlg
* *
Ch4*1149141 SflO*40C400. ..Ffl*M.S4lflTo0aVl 0190.10*04*4
1919
F'_2Q4n
14.0
Ç94994049
Jcp.1-c400*,4r.00 519 0*. 199
000*1400* 941*9.0*09 944 24102171111 *1*99*41
140*1*01
4*11
Co,
9*11,71
040O
ww#.Cfl.P1Ifl0*MiS0*0W.7CIt4M
10019 4910s04*09400i 01.lpl74ay!
*Jtn
000*40*4
094 *400*I70S4*1Ofl 40° 9(Ø9*19949
209*02*94*09*4
Ml
.0-o.,!,
*9*10*09*0.2
$1
00*1016* 110000flça01211641 19&.448*14*9094169 90h&h9w*ls0fl0i40
241
!00Ç*79!00W911u4Ml Q4S71O715f9409999sQo4**0
4h411Cl0.M.4* .44*144, 0111900 Ø714 411
.1*4l4 (*4*9017 *0*. SlSpp$
*99*150 F4144**99400*1*l4$h**4.*A1t4M01l10*O*J0*I1
.IoppIl997**4,*h.y**ll*0****
*
*0*04*09 *44*74 94*416* 440*10 9,00*400*94*9.19*4
2híMlI0299000
afl,7990q**l1.7'n.10941*nl 440°
19*
*t990*o41SS004l47,
lfll***tl000*0*
14*4*4
4*0*19
41l.41
4;,
9(0111*400*00*0*
9.004*000*1110 !,*0*&9 4* 1 00I*I4 *4OO*0W*900024 91°
4
*
4*.*19
$40*1
*919*01
*4*19
*000*04.197 10*1610 9*70*4*4*4*1*0*
.74
WC
00,**l*o0l
•
9( 1
4,9Mlhboo*.,. 19*0,4*4199.4*40 ,,hl0900*1S0104fl900*77 0000400, 4,
2*00*4*440 2*0041100*0t*4V0(0214000m** 70*41*0 044 4104*41. '0'0.
IOIOØ
$0040
1!?
14*002
149404
50007$, *97! 9(0944*4 4*0* 0*1096*191
7hoçlflSç.4 2190404*, 0*140 0*014,9*.,, C,W1**019. ro**100
35.
0*44400*0 *0*0*0*,
14*00*461*191
9940t99Lli992)
*0*4*0 0*07*4
.o000*0400
04*4*
0*000*64*4
*
O próprio Bing, em seu site, reconhece a importância de fornecer respostas rápidas,
e acredita que os seus anúncios de produtos beneficiam os usuários por essa razão: "[users] save
time because theji no longer need to navigate to a dedicated shopping experience tofind what the,y 're loolein,g for. ''
36.
A explicação para esse desenvolvimento é simples e análogo às seções de anúncios
de um jornal (slide 39). Apesar de os jornais aceitarem diferentes tipos de anúncios, dentro de um
mesmo jornal existem diferentes seções que são especificamente dedicadas a certos negócios,
oferecendo determinados produtos e serviços. Se um jornal fosse obrigado, em uma seção
específica, a exibir anúncios de outros jornais mais especializados, os leitores certamente teriam
uma experiência pobre de leitura. Para ilustrar, seria confuso para os leitores se, em uma seção de
classificados de um jornal, especificamente dedicada a anúncios de carros novos e usados, fossem
exibidos anúncios direcionando os leitores desse jornal para outras publicações que também
exibem anúncios relacionados a automóveis. De modo similar, o Google exibe diferentes tipos de
anúncios na sua página, incluindo os PLAs que são limitados para varejistas que vendem certos
produtos. Notavelmente, assim como os PLAs do Google, o Buscapé não exibe resultados de
outros comparadores de preço dentro dos seus próprios resultados.
1
http: / /advertise.bingads.microsoft.com/en-us/blogpost/1 27643 /bing-adsb1og/bing-introduces-product-scarchand-new-ad-formats-what-it-means.for-bing-ads-advertisers. Acesso em 05 de junho de 2014.
14
p.D
37.
Por fim, do ponto de vista dos varejistas, a inclusão de sites de comparação de
preços nos PLAs também não faz sentido (slide 40). Permitir que sites de comparação de preços
comprem PLAs levaria a uma duplicação com as ofertas anunciadas pelos varejistas. Os sites de
comparação de preço não vendem produtos próprios, mas são intermediários que exibem ofertas
de varejistas e recebem pagamento quando os usuários cicam nos sites dos varejistas.
4 044,00.4
Os varejistas não querem um de seus
parceiros oferecendo seus próprios anúncios
de preços em outro site também parceiro.
11.
40-40000 L.,,,wo 5404.044100048* e 05400,4403 40550300 LF0 340 ll',44300.Jven 0õ'4040s 03440
sk.4,'S'Ç3314 3 '. oSl
40138
Go
gle
140004Q0503
403
. •
j
14(0.0044 1s0000 4000. *44!00040034"4,4,30404(t 0
NIO(OOOIO ldenpod 8400 omo33çonas conoto
04
LeStOS 5400
£
100840404000044140. *44! 109008*0484
14080034
4444104444 co..
IQlt.o'Leoo,'o
444 O
444 40, 0,7wç.4 43 04,
FO0IO000S3(JO
40020000403444
304100054000030*4,04
E00o353 PeifojI;04!oetoo
103000 5405.08A1000GSRCoflInlOI 0003
2 4*0144(38 LEO 0014
osi.sya,,v tocor,c
.,1.538
'
NolebOOS L*novo Preços on Ruscape
-
«1401 800830$ OVO 00, (000040400a
OIIMO44IIL000 0440 40.41404000 0
040,S4034444404W 4_O'
'
304
0 44
OCO
Isnovo 14' deaPad 8400 ColocO N0300ool, RO00ow
04040 jfl08q 0o''*,o'ct'.*0n670'0244
'.
D03013T110 Tape BOOJIOO7
004144,0$.,
-
'
o
38.
Além disso, é natural que os varejistas não queiram que os sites de comparação de
preço participem nos PLAs, haja vista os efeitos previsivelmente negativos que causariam sobre o
custo dos PLAs. A inclusão de sites de comparação de preço elevaria o custo das ofertas dos
varejistas, e não porque eles estariam exibindo ofertas de produtos únicos, mas porque suas
ofertas de produtos são, de fato, as mesmas ofertas anunciadas pelos próprios varejistas. A esse
respeito, deve ser frisado que os sites de comparação de preço meramente exibem os produtos
anunciados pelos varejistas em seus sites. Sendo assim, os varejistas acabariam competindo, nos
leilões do Google, contra os seus próprios produtos anunciados nos PLAs.
15
39.
Os screenshots abaixo demonstram que os PLAs incluem muitos dos mesmos
varejistas - oferecendo os mesmos produtos e ao mesmo preço - apresentados pelo Buscapé,
Bondfaro e outros sites de comparação de preço (vide abaixo - slide 41).
5.3~ 5r3g
SMGO)M
Resultados no Goo9Io $floppng poro
t
S.mMrng
OS22.t
Sarno,.,,
P011415
450.00?
Sm,t.
rL1
S.3s.ng
CM.ay $5
RS2.G1P.41
2.0045510
CO4s,
S..rnflphaa*
05210.12
0450,4502
Sn.lpb
G&to,y 55
012 195.00
510021
I
092.199.22
—11
o
o,
2.287.,
.512.222,.,
40.
É importante observar, no entanto, que qualquer site de comparação de preço,
inclusive o Buscapé, Bondfaro, Zura, ou outro controlado pela E-Cornmerce, são bem-vindos
para participar dos PLAs referentes aos produtos que eles escolham vender diretamente aos
usuários nos seus próprios sites. O Google entende que o Buscapé, para aqueles produtos que ele
vende em seu próprio site, atua como um varejista (vide abaixo - slide 27).
16
Srnitphone App'e iPhone55 16GB Destoqueado
****
Ri..96S55 R$2.894408
Eco
P:gç
Ct.h4n Si
P,odts
4444Ç
{
Avohaç*ifl do, Co4rn4dft8
otoi
Vd,*4
t'
a - ;4,
4
à
RS 2.626,97
-
Coco*
A política do Google que limita os PLAs apenas para sites que vendam produtos
41.
diretamente aos usuários não foi desenhada para excluir competidores. Varejistas como
MercadoLivre, Americanas.com, Submarino, Extra, entre outros comerciantes, também
concorrem diretamente com o Google por usuários que estão procurando informações sobre
produtos. Ainda assim, esses concorrentes são bem-vindos para participar dos PLAs. Claramente,
•
a política do Google de tornar disponível os PLAs apenas para sites que comercializem produtos
diretamente para os usuários possui um propósito legítimo e razoável - a saber, proporcionar
uma experiência de compra mais rápida e conveniente, por meio da redução do número de
passos na busca de um produto - e não tem o condão de prejudicar sites de comparação de
preços.
E. Conclusões
42.
Com base no acima exposto, e tendo em vista os elementos apresentados pelo
Google na reunião havida em 28 de maio de 2014, é possível expor as seguintes conclusões:
17
GR [NB ERG
CO1.DOViL
o
1
.ADVOGADOS
•
As ferramentas de busca começaram a exibir resultados temáticos, mesmo no Brasil,
muito antes do lançamento do Google Shopping;
•
Os resultados temáticos fornecem respostas diretas, rápidas e convenientes para os
usuários - eliminando passos adicionais de pesquisa. O consenso geral é que exibir links
para sites de comparação de preços dentro do Google Shopping proporcionaria uma
experiência de busca ruim dos usuários;
•
• O Google Shopping beneficia os usuários ao fornecer resultados diretos
(independentemente do fato de serem exibidos nos resultados de busca orgânica ou como
anúncios);
o A transição do Google Shopping de um modelo gratuito para um modelo comercial
beneficia varejistas e usuários - uni jogo de "ganha/ganha";
•
O Google Shopping nunca foi projetado para prejudicar sites de comparação de preços:
(i)
O Google continua enviando tráfego para sites de comparação de preço;
(ii)
O Google continua vendendo anúncios de texto para sites de comparação de
preço - esses anúncios são ainda mais proeminentes que os links do cabeçalho
do Google
•
(iii)
O Buscapé, Bondfaro e qualquer outro site de comparação de preços é bemvindo para comprar PLAs, desde que comercializem os produtos anunciados
diretamente em seus sites;
(iv)
Desde o lançamento do Google Shopping em 2011, no Brasil, o Google sempre
permitiu que varejistas que concorrem diretamente com o Google participem,
incluindo MercadoLivre, Americanas e Submarino;
18
Fis
43.
Ratificando os termos de sua defesa, e com base em todos os argumentos já
apresentados, requer que o presente processo administrativo seja arquivado, no mérito, diante da
constatação de inexistência de qualquer conduta anticompetitiva praticada pelo Google.
Termos em que,
P. deferimento.
São Paulo, 19 de agosto de 2014.
~
Cordoã
.à&
OAB/SP 233.058
OAB/SP 290.059
oaatest
OAB/SP 154.609
19
o
ví
4:
ria
(R
=E74'o
0
cz
a
o
-
co
L)
0
•
.-
co (1)
>
a)
a
03
c,)
(1)
co
0
O 0
C/)
O
O
o
CO tc0a)
CL
0)
O
a)
E
co (D
•
Q
* = (1)
LL CO
co
o
O
O
o o
0)
00)
cco
U) CL
co
co
(1)
U)
Ocz
a)
to
0) (1)
0)
o E o ""a
O
O
O > O O O
O) OWOOQF—O
•
•
•
•
•
•
•
o,
-oo
o
:3
0-
0)
o
co 0)
a) o_
ci
co
0U)
OCO
CO
ç0)
o
00
(1)•-
oo
CO
a)O
M.
(1)
a)
00
-o:
a)
0-0
:3cl)
oocDwo
a)t, 04a) >C
"
L..
wi
oca) 0-c0
-o
><
cOc°>
a) -
4-,
c
CJ)
e
• •
e
.l.
%-i
EW
S
o
o
o
Q)
(1)
o
o
(1)
o
o
o
—
—
-
o
co
0
0)
—
o
o
CD
o
o
o
cDOO
co
CL
o
CD
(~5c&
-
/)
1
•- AW
•
CI)
•
'ti
E
.
VI)
0
Lu
ci)
cl)
L.
cDo
-o-
(1)
(1)
-coa)
CI) °
a) 0
QN o
(o C13
•
0D
co
o co
W c0
0*-'
E D, 0
772
0•—
o
co
Cl)
)
-I-J
cz
=
CD
c1)0E
CO
co
Q)
c:
o
(1)
CO
.EEOO
=
°cnE
cn co
000U)
Eco°
Xcoco
(1)0
cn (1)
—E
(1)
0Q LC',
a)
o
'cooa)
E
- =0
o
-o
COcO
co
o
co
co
e
co
.
e
Z15
uI
ç.
2•
o
'1
1
foi
.-
=
,
21
-
o
r
•
t
,
;
:
ri
E
•i!
=L-
Lfl
=_)
'c
d
E •
-. .,
í4
-.
ci.
Ci
'2
.
O '
L
r. 1
-.......
—.
2
-
o
d
o
=1
C/)
o
>•
!
...l
-
2
I
4)
ni
Ot
(is
-'
ip
-"
rL
-
,
*-
LL
-
-
-
w
-
C
-
.
:
c
2
o
c
o
*
-
4-2
fu
O4LC
Qoc
,-
o
í1
4
i
...-
;
4
1.
-,
-
mil
1
,-
t'
4
-
444
4:)
ii
III
Gi
V)
o
•
4:4
o
o)
o
1
fl
s
LI,J
o
*
I
co
CD
(1)0)
00
0
coo
o
0>
E
00
e
gm
>
O)
Q3
q
COQ
8
o
0
O.!2
•
o
«o
0)
co
.0)
CD>
Eu)
COO
«o
co
Co cA
co
O
co
o
00
-
O)
0
:
•
O)
j30)
0)C/)
co
c: u-
4-'
0)
• cá)
(>
O
lo co
O
0)0
G)
E0) 0
_
-(1)
o
cACO
(Du)
o
00(0
CO
O
F
m°
Ci-
0u)
(O
(1)
E
O
o 0) 00)
oO)
CO o
N
O0 •
0 co w .
1
0
0
E
o
QU)O
.S2
Co
cm
0
')G)
E
co
cz
—
ui'
c
O
'-O
•
0
C1)
O
a)cl)
co
co
0c
=
O
O
co
E -0
C13
O
O
-Q
U)
=0
ot
oco
C/)
0-U)
C/)
w
0
CO
E
L
O
(1)
(1)
4-
o
E
E
o
cl) cn
COCO
-Q
E
0
-o
-o
W
O
o
co
-
.S20
>cDc
CON(I)
o
o
-o
S._ a.
o
(Da)
-o -.-'
g -o
-o
r
o
Fa
O
E
COO
0 0
(/)0 O
O
w- c
c o
=
.
OOQ0<
'
-g.h2
2
cn
p
LD)
4J
05cD
ogo
3DU)
U)
e
cDDU)
a)OCO
a)0
LÜo
= Q <a)Oa)Cfla)O
coa)<oc,)S -D •C
a)r_-c\Ja)
4CO<c
E
O)
!
o
o
5
<
ci< o E
O
:[
H.
co
Lo
-F-
.Q
it
E
•2°
>(i)OW
>Ci)L)
2
o
c
o
o
o
o
:2
2
tc
O
E
o
(o
o)
o
cri
E
:
o
=
O
O
co
cD
O)
O)
C'J
cs
O
o
oç
02
EE
.o
.
Eo
.ca°E
'E
-
Ea
Oc
' '
o9
°•
20
E'
mo
2'o ''m
L4
o
oO.Qo
CE
- -Ou
0— '
LLO
t
CO
-om
co
vi
o .!2 o
-
'
o
E-ao .-.
I
o—
2
2
L
uo
.
(d
-E
CO
O)
E
2
:,o&
- .o -
=
!I
O
Ó5
m
-E
o
o E -5c'
2c
o
-2E
cnCOa.i-
çr
J1
L1
o
.
11
Z
(1)
t
o
Co.)
:
-
(o
.S2
o
cI.
(1)
o
-O
O)
o
=
o
(1)
o.
o
o
(!)
ca
a)
o-
-O
(1)
C/)
o
o
_
c-
irI1
r
c,
J4
-
-1
tV
ci,
tJ)
c
Cii
(
(r
r
l
=
c
ci,
j
LU
KN
i
ci:
ti
1
-
F
E!
Li
5
L)
o
rui
i)
1:, n
l)
—
CL
O
c :
ti
M
o
ti
tti
«
Lii
-
Ui
r
-
1
D---
- -
ç:
-'i
---'-
Fts
rr
W
o
-o
5
L)
:
à
1l
c:
1•
:;I-
a
—
£
-—
E
IL
ct
LiJ
-
lfl
CL -1
J)
tQIJLJ
1k
1)U)
CMr
o-
1']
-
=
o
w
:
-J
Li
CL
E
-
LII
*1'
—
t
t
EI
E
1
WI L
-
(1)
-
1-
C2 -
IZ
c_I
c
t
E
1
—
co
>
Iço
-
t/l
o01
li
r
4)
C)
',
4)
0'
IILL
a
E
'1)
1
8)
.
E
•'
1
O
•
2
W
O..
,
•i
,i
,
t_I
O
••
O..
•
j
-
,
-
r
'i
(3
LI
A'
LLJ
'aMS
-•
•
E
oi'-' "3
A'ti) L'
- 2 2i
'13
-,
1 •
-J
8)
it..'
CL. 2•
'
'3
a':
E
153
'
-t,r''
i
LLI
..
C.
• _
A'
Li
O
•
't
0'
'8)
CL'
«
O• •, ,,i.
t/l
n
•,
(1)
0
(
00
0
-D
co
—ci
o
-o
o
0 00
(1)
. O..o)D-o) 0 = E co 0..t
O ()
o ci,
,-'
-a
2
9-
Q
c
(S
L)Cfl>LO..
(1)
o CL
.4..J
0
co
o
9o
00
c
L'O
o
o.. -o o.a1 o
Co
tu
0
o
Co
tu
<a)
-a)
(1)
J
a)
•
O
Cm
CÜ
E
cao
c
co
(1)
0)
E
0 <Co
>
c 2 2'E
o
0O
O
Co
«oco
0 0 CO 0
E
c co,_
a)
U)
- -
o o E E E
O
Ü 0)0
O O
cc
1
(O
co
(1)
o
Cm
O
(1)
w
a)
O
(I!
O
(O0 cO
)'
CD CD
CD Q
CN
•
o
L_
CN O...
.
0o
a)
2D
Q 0 0
Co
-g
c
co co
C
D
>
<
c
<
c oci.
(O
U)
E
0
CL
c 3 (')
2o o
(•%
- O '-'E
c/)
O ài
iE O
CD
p
CNI
•
- - o
CD CD
CNI
cm
•
•
a)
Li..
•
—a
.
FIS
L
co
•
..5
E
0<
OD
o
> cl)
-J
-
O
co
co
co
00
CE
o
-
00
N
rO
<co
o
o
o E
C
CD
o
o
CL
Oo
's
o
0)0
II II'
•
E
0)
-
Q
1
-
O
-U)
(Oc
CL
4-
-o
E
o
o
ao
o
c
<cOE
CS)
e
o
O8
tPt i:
z,.___
U)
1X
N
E
co
o
CU
ri)
CO'
-'
O
co
(1)
O
Ct
uJ
oIot
CL
6OO
6001
1
1 ;
j
r
goo
,00ç
1
3
1
LOO
J
9OO
UM
:i<
;;
.;•;
çooi
N
TO
1
As
o $
3
—o
(1)
eo
fl;
rt:
O
w
o
o
ci'
rn
ÇD
<o
b
O.
O
.c
o
w
o
co
Or
CL
CD
e -o
(l)C
r.
w
fiL
E
ct,
c,)
•
•o
CD oc,)
o<cs
CL
E
E
m
ou)
o
o
o
1'3
o.
-cO
o5
E
2c
.= E2
OODO
N
o
o
o
0
O
G)
o
CL
o
a)
=5
c/)c:
C%3()
o
5=c
oS< c\i
E
(
CI)
()
Qo
E
o
O
OL
oI1
cfS
U)co
Wo..
c
O>
1
:
E
(1)
*
o
E
a
)
1
câ
WL
—
• E• —_
LJ
-
411
(1)
•
U
ci
1
!
1
W
rç
Z1
•
(III)
co
O
til
1.
Ao
4
_
Li.
lw-
kh
04
•ii LI
I
:
•
• •
-
.i,.
14
V
o
CL
rr
><•
W,)
cu cn
oc
zE.
L
'1lii
:
•
Z(T
•'
1
as
li
—
E
a
•
—(a)
E
O
o
O
o
cl)a)
O
>
-2
E
N
ai
40
CL
cI)c
°G)
—
a3
C3
'au
'l)
G)
O
O
O
G)
Icr
o-
O
COCO
E 5 9°
ci)c
-
aia
o8<
(1)
o
11
0
o
co
fl
C1)
CIO
00=
•—
L
—
CO o
a)
C
Ci)
o
EtQ
=
COcj)
cC/)
cI)c:
(o
o
o co co
"
CO
CO
co
Co
cOCO
E0
C/)
oo
CI)
O-CO 0
--=
-
a)
Co
-;-;'¼Jr
ro
'
oco
-co
co
L.7 CO
0
COCO
5uw.2
-
II-
a)
Z C)- -0 -0
C1)
a)
-
o
-o
E
o
ØL..
-oa
a)
3)
.0
0
o
Cl
0<
E
o<
U)
0
•
E
E
o
O
co
0
>
0
o
-o
E
o
o
o
C1
OciDO
N
O
0a
-o
<
G)
D
oa)
o
a)
0
o
o
E
co
E
a)
-o
o
o
O
O
CD
cr
O
CL
°
o-
C/)
00
a)
OC
•
•
CI)
08
< -
E
>
0>0
C)
CI)
CD
E
a)
o
0
=
-o
o
o-
(1)
o_a-
(1)
C
=
a)
-E
>
=o
o
Eu
co —
>1
E
-
CDC
cDo —
rnc,)
k.
8
o
(1)
>
Ci)
o
ci)
-
CD
CI)
co
CD
()
cnE
(1)
E
0(/)
0W
U) 5
(1)
•
(1)
>E.
L
COC(J)
OcOcD
o
Ci)
W
o(1)CO
a)
E—
cr
E
o
EwaH
0>
OwO.
—O
.
E°
c E0
ZM
> O (J)CD
N
(1)
0—
E
(1)
2 —Ci)
0°
2
a--CD
Q
Q
2o- °0 o
O) E
O)
"-a
o
o
: (1)
ci)
C00 Da))
o
=
0
EOEE
ECO
.
CD
o
0
O
• CI)
0
.
(D 0
E
cri
D
d
(1)
o
c
uj
(1)
(I)
E°
(f)O
ML
04
CO
O)
CO
c:
CO
20
o
co
CL
L.
oco
cri
03
Ou)
o
a)
5
-o
(O
o
-o
o
0
03
a)
"O
(O
OcT
cri
>
co
z
-'
ça
'
o
CO
Cu
o
EL1
J
-o
ri
- + co—
o
o
cD
o
a.
o
P
IM
CL
-
(1)
-
o
o
-
O
5-
-o
o
LI
li)
U)
0 g 0
(1)
o
L,
ai
Dc
U)
CO
C,E
a)
•30
c0
00õ
oE.
><-0
0
- 0
0c
E
0 0
0U)
E°
Eo
E°-
0
rL_c
.-J
(0
L
.
o- ujw
•
0U)
(1)
(1)
o
.0
•
%—
CO o
(1)
O
co
0
.
W
0-'
E.EcL
3c0
o
°EE
0i
M
0cU)
00
E 0
j o 5cE
—c0-
2 (0 o
00
•
D
a
C
a
W
W
o
Q
(1)
E
E
0
00
2°
'-,'°
()
E3o
0/)
-_L.
0c
0=
(0($
c
.2oc
C1)
cl)
LI)
C0
U) m
a)
LI)
Z3
Eo
Oco
Q_a3c, = (1)
E°0
0 o
(1)
0(D 9— O
09-E
o
—
•
0
C/)0
<E ° 1
oE
00
=
000 (90
0(l) (1)
O0O •
0 -0
•
.
•
E
o
a)
a)
a-
o
E
o
o
4;
7
Q)
4
4787448
44
7
4444
4
1
k!)trJ
443
21
(1)
4
AfN1
44
4
344444
a.
.2
E 0
O
rn
. _
I-J
P%
o
.
2cn
O
CO
(1)
84
CI)j
4
4744
444
," co
4
44
847 44
4447
7
W
4
47
za
4744 44
b
44
t4
8»
-
IN
-
$44
8
cu
(1)
te
44
744
(j
:74
44
4
4
4
si.
c.j.
co
CL
Ou)
rn
+i
a)O
a) a)
*
1U!
'.
1
E
co co
CO
C,)
ii
W(/) t7j )G
o
. C.)
ca
7
2
.':i 144
84
co
4
o.
L2U)
.o
•4lW
O ..0
CQ
O'Ci)
Jsl
O
CY)t
ci
4
Qi11FI1
44 4
Ç?
(j)
84
0.
4747
2
ci
O
1
44
•
44
o
0
0
L
oa)
-o
(o
O)
1
c0Cflu
>0
-Q
01
ti)
a)
-o
O)
(1)
• 11
o c.
U)
(00
D
b
E Lj::
3c1)
a-
0o__
o1
5I
DI
1
c0
E
DOL
cflO 0 U)
D
co
00
•Cl)
-
__q2
U) U)
o
-,
C1)
o
o
CO
0-c
L
ç0
=c
cü
0(0
(DSj
CM (1)
0(1)
cL3
-C1))
a)
W()
0(/)
3(><
<coE
o
•o._.l
C2
0u) COcO -E
o
0 U)ti
W 3 o
D0 < a)
0 -0
o
c
00
co
E
E
o
0
E
g
> 0)
3?
0
0 0
(O
(1)
- Ec
a)g
5EEcco
COO
a)'5
(/)a <o
e
e
e
4
rã
N-i
A'
ÍD
L
tf
tj
1I
-
pj
Vo
co
0
O
o
c
tt
o
(I)
(1)
o
6
cii
o
O
O
Ci)
O
O
o
>
0<
CL
Ecn
O-ci
OQ
_
E
E -O
Ou,
Oc
(O
-
* c-g
Eo
o
C1)
•0
O C)
:
CO
O)
cE
Ci)
O
O2
o
EE
Eo
20
-o
c,,c/)U)
O0
cED
°Eo
=
O
O—O_
OO
cO=
/)
o
(OCo
=o
EOco
O
CO
co
C/)
•=(/)0
o
oP'
-o
cn — — °E
O
O
.
w
O
O
co 0
—:tf
E
=
00u,
Cl)o
(i)CI)
00
o
O
E =
o
O
u,O
O-CO
COC/)
E
co U) M
co 0O
.
.
O
O
0 = 0
cn
o
-I-J
Q.
02°2
o
LI cn
mo
ç1u
—3
-
gt
gt
-J
LU
çp
LU
-O)
P
3-4
ti
bl
O)
tu
(
1
1w
)R
z
LU
Y
z
'li
o 1-
z
<Lu
92
z
uJ
1-
•
2
w
O)
iu0
o
o
o
w
1-
.—
z
00
LLI
2
z
C)
00
w
E
o
II)
E
1
z
-3-
—4
O
lJJ
-
00
004
('4
en
1
1
00
CD
CD
001
en
m4
1
cn
cn
1
enn
('4
N
001
('4
1
—4
o
O)
L)
1
'o
Me
O)
O)
-
O)ri)
o
a
ri)
.HP
o
a
C/D
-O
-c
ce
•
Cd
ce
o
o
1
o
o
—4
•o
N
z
wil
o
-Cd
-
t1D
o
z
40.
o
02
-
o
ce
o
j
o
Yedda Beatriz G. . . . S.
Singer Afonso
DAS- PGG
2.2 CGM,
VERSÃO PÚBLICA
ILUSTRÍSSIMO SENHOR MARCELO NUNES DE OLIVEIRA, COORDENADORGERAL DE ANÁLISE ANTITRUSTE DA SUPERINTENDÊNCIA GERAL DO
CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA - CADE
As
CADE/MJ
Protocolo - Geral
N
08700.006951/2014-10
-
E;?
Processo Administrativo n. 08700.010483/2011-94
GOOGLE BRASIL INTERNET LTDA. ("Representado"), já qualificado nos
autos do presente processo administrativo, por seus advogados infra-assinados, vem,
respeitosamente, em atenção ao Ofício 2316/2014/CADE/SG/GAB, apresentar resposta à
questão 6 formulada por essa Superintendência no Anexo 1 - Questionário, como segue:
1
1 I\IV'L)1.L1'.J
rORD0VIL
UMYLIIt71 II*1
/ •
VERSÃO PIIBLICA
Anexo E - Questionário
6. Informe quais são os maiores anunciantes do Google, no Brasil, nos anos de 2010 a
2013, classificados conforme receita do Adwords nas seguintes posições:
10 ao 200
(i)
(ii)
100 ao 120°
(iii)
500 ao 520°
(iv)
1000° ao 10200
Adicionalmente, informe os endereços, telefones e pessoas de contato para cada
um dos anunciantes informados.
/
2
t
VERSÃO PÚBLICA
Finalmente, o Google requer que sua resposta, e, particularmente, a planilha em anexo, sejam
tratadas como de acesso restrito, para acesso exclusivo ao Google e ao CADE, com
fundamento nos arts. 50 e 53 do Regimento Interno do CADE, haja vista conterem dados
confidenciais e de propriedade do Google e a natureza comercialmente sensível de uma lista de
clientes. Nesses termos, requer que nenhuma informação apresentada seja divulgada sem a
autorização prévia do Google.
Termos em que,
P. deferimento.
São Paulo, 22 de agosto de 2014.
•
-O~nCordávU
r ~
-1~oMotta~
~
OAB/SP 233.058
OAB/SP 290.059
OAB/SP 154.609
3
(!fÔJ1
GRTNB
ADVOGADOS
BeatÉ À. 0. C .
VERSÃO PUBLICA
Modli
EPPGG
S-2.2C
kD
ILUSTRÍSSIMO SENHOR MARCELO NUNES DE OLIVEIRA, COORDENAD
GERAL DE ANÁLISE ANTITRUSTE DA SUPERINTENDÊNCIA GERAL
CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA - CADE
CADE/MJ
Protocolo - Geral
08700.007313/2014-17
2
Processo Administrativo n. 08012.010483/2011-94
GOOGLE BRASIL INTERNET LTDA. ("Representado"), já qualificado nos
autos do presente processo administrativo, por seus advogados infra-assinados, vem,
respeitosamente, em atenção ao Oficio 2316/2014/GADE/SG/GAB, apresentar complemento à
resposta à questão 6 formulada por essa Superintendência no Anexo 1 - Questionário,
protocolada em 22 de agosto de 2014/
1
,
GRINBERG
ORDML
ADVOGADOS
Finalmente, o Google requer que sua resposta seja tratada como de acesso restrito, para
acesso exclusivo ao Google e ao CADE com fundamento nos arts. 50 e 53 do Regimento
Interno do CAI)E, pois conte dados confidenciais
'
e de propriedade do Google e a natureza
comercialmente sensível de uma lista de clientes. Nesses termos, requer que nenhuma informação
apresentada seja divulgada sem a autorização prévia do Google.
19
Termos em que,
P. deferimento.
São Paulo, 04 de setembro de 2014.
/Í.R.TLumi
/TTordovil
OAB/SP 233.058
()AB/SP 343.057
LY1
Malatesta
OAB/SP 154.609
9
/w.
0•
N,
■,...-\
,.
_ ,. . . .i
,,.
SN.
‘1‘
-
4)
L.,
›
:
O
J
cli
-ri
cks
c.)
cli
no
czt
c)
o
O
o
0
cri
(1)
›'
(1)
C
Cd
1.-■
Cki
cks
Z
<1)
›
<1)
C
ni
O
01
i•-■
.O
É
o
c)
cd
_e
01
b
@.)
"C
I
3
44
O
0
d
-e)
c.)
k
ah
„....._ 4.
bi
e)
t
‘.....1
/
Q.)
O
....._„:,_.
,....3
•
Z
-r-s.
00
00
oo
oo
C•1
1
.--•
C•1
en
en
i
.--•
■
C)
.._
C4
p.,
•
r4
,„. ,
,--4
C•1
en
en
1
,--4
■
C)
...
e ,,s
C
o L)
o
o
o
Lr)
o
L9
o
Lr)
o
Lr).
It:
CO
CO
1
1—
1—
CO
Lr)
o
toe
F.:
co
C?
1—
1—
C?
-,d,,
.
›o
-cs
o
,
.
o
Ti›,
-tC
-tC
o
1—
1—
0
^ri
o
Lr)
o
toe
co
(;)
CQ
1
1—
1—
iNn
........
%Sb
0
'...s.t
cr
^c)
obi)
,
Ti
.<
obi)
,
-c)
.<
O(1)
.c_
•cp
----
C.)
o
c.)
c.)
c.)
o
çã+
o
cks
s...
<11
>.,..,
3
o
c.)
o
•zr
to
Ti
v)
O
c.)
o
ID
ID
Z
o
cl)
c)
N
=
O
cz;
cl)
(2:1
czt
-ri
-ri
cl)
C.)
'3
c
o
o
›-4
O
_J
Cd
-6
2
o
12
cr)
c)
.C.E
22
Cd
_J
-cr)3
_E
O
..-
17;c
15
=
X
c
2
—
c
5LL.
,.......e„.
--....4
4
..,,.
....i)
\....41
CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA - CADE
1
CADE/MJ
Protocolo - Geral
/
08700.010315/2014-93
g
Processo Administrativo n. 08012.010483/2011-94
ré
•
9 -.
GOOGLE BRASIL INTERNET LTDA. ("Representado"), já qualificado
:
nos. lautos do presente processo administrativo, por seus advogados infra-assinados, vem,
resitosamente, apresentar o que segue.
Em 22 de agosto de 2014, o Google apresentou à Superintendência uma
planilha contendo informações de contato de anunciantes, no Brasil, referentes aos anos de 2010
a 2Q13, de acordo com a classificação solicitada por essa Superintendência.
/
V1•:.RSÃ(I)
Úm:ic
1
II
VERSÃO PÚBLICA
Finalmente, e considerando que as informações ora apresentadas são de
natureza confidencial (lista de clientes), o Google requer seja conferido tratamento de acesso
restrito, sendo seu ACESSO EXCLUSIVO AO GOOGLE E AO CADE; nos termos dos artigos 50 e 53
do Regimento Interno do CADE.
Termos em que,
P. deferimento.
São Paulo, 1° de dezembro de 2014.
fr1en
~
CoJovil
OAB/SP 233.058
OAB/SP 290.059
ú+%eoTokuzumi
OAB/SP 343.057
atest
OAB/SP 154.609
VERSÃ(I) P("BiJCA.
2
VERSÃO PÚBLICA
DOC. 01
[ACESSO EXCLUSIVO]
IRSÃ() PÚBL.E
3
1
ualiza, dø poc
¼
Dados dpiocedmenI
H
MINISTÉRIO DA JUSTIÇA
CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA - CADE
SEPN Conjunto D, Lote 04 - CEP 70.770-504 - Brasília-DF
Termo de Encerramento de Trâmite Físico
Processo Administrativo n° 08012.0010483/2011-94
Representante: E-Commerce Media Group Informação e Tecnologia Ltda.
Representadas: Google Brasil Internet Ltda.
Natureza: PÚBLICA
1. O processo em epígrafe foi devidamente convertido do suporte físico
para eletrônico no SEI, em conformidade com o disposto no art. 16 da
Resolução n° 11, de 24 de novembro de 2014, mantendo o mesmo
número do processo físico (NUP) e mesmo interessado.
2. Foi efetivada marcação da referida conversão no cadastro do
processo no SisCade/Sistema Processual e que o processo físico será
imediatamente encaminhado para o Arquivo Geral.
3. Fica encerrada a tramitação do processo em suporte físico, sendo
vedada qualquer juntada física de novos documentos, para, a partir de
então, ter continuidade de sua instrução e tramitação somente por
meio do SEI.
4. Para fins de registro, o processo originalmente em suporte físico era
composto de:
1. Folhas: 2114
2. Volume: 8
Apartados: 13
3. Mídias: 11
4.
S. Unidade de lotação do servidor responsável pela conversão, signatário
do presente Termo: ProSG
6. Data na qual se deu a conclusão do procedimento de conversão:
31/12/2014.
Brasília, 31/12/2014