demand - Columbia Institute for Tele-Information

Transcription

demand - Columbia Institute for Tele-Information
Demand
y
Analysis
For Media &
Information
Products
© Eli M. Noam, October 30, 2010
1
Start of Lecture
21
1
The Media Value Chain
Resources:
HR
Finance
Tech
Accounting
of
Performance
Value
Creation:
Strategy
Environment:
Production
Marketing
IP Creation
Pricing
Info.
Environment
Distribution
Law &
Regulation
Demand
22
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
IV. DEMAND EXPERIMENTS
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
• Is This What Media Firms23Need?
2
I: Why Demand
Analysis
24
http://www.sunways-direct.com/magnifying%20glass.JPG
• In the previous chapter, we
concluded that one of the
characteristics of media
companies is the high risk,
uncertainty and instability of
demand for their pproducts
25
3
A famous Hollywood saying:
“Nobody knows Anything”
- William Goldman,, ((Columbia
MA ’56) Oscar-winning screenwriter
- ((Butch Cassidyy and the Sundance Kid;
All the President’s Men); Stepford
Wives, The Great Waldo Pepper;
Marathon Man; A Bridge Too Far; etc.
26
William
Goldman
“Nobody
Knows
Anything.”
http://images.google.com/imgres?imgurl=http://www.wga.org/uploadedImages/news_and_events/101_screenplay/goldman_william.jpg&imgrefurl=http://www.
wga.org/subpage_newsevents.aspx%3Fid%3D1679&h=1525&w=1500&sz=1278&hl=en&start=12&tbnid=6TCls5WzVoSMtM:&tbnh=150&tbnw=148&prev=/i
mages%3Fq%3DWilliam%2BGoldman%26svnum%3D10%26hl%3Den%26lr%3D
27
4
The Question now is:
• Is Goldman right?
– Does one really “never know
anything?”
• Or, more correctly, can one
know better?
• Can one increase the probability
of being right?
28
29
5
Case Discussion:
A
Hypothetical
h i l
Case
30
http://www.bestchoicecare.com/library/images/tvcouple.jpg
Case Discussion:
“Viacom Golden Years
Media”
• Viacom is considering to enter the retirementretirement
age market
– Through multiple platforms:
¾Cable Channel
(“Golden Years Channel”)
¾DVD (“B
( Bestt off Golden
G ld Years
Y
”))
¾Magazine (“Golden Years”)
¾Website (“GY Portal”)
31
http://www.bestchoicecare.co
m/library/images/tvcouple.jpg
6
How would Viacom estimate and
measure its audience, their content
preferences, their consumption
preferences and their willingness to pay?
http://www.cdc.gov/communication/images/tv2.jpg
32
http://www.outsidein.co.uk/photos/sunray%20watching%20TV.jpg
Viacom’s Existing Cable
Channels
• Ordered by target audience age
– Noggin (pre-schoolers)
– Nickelodeon (tweens)
– The N (teens)
– MTV, MTV2 (15+)
– mtvU (college)
33
7
Viacom’s Existing Cable
y target
g age)
g )
Channels ((by
– VH1 (25+)
– Comedy Central (20+)
– Spike
S ik TV (30+)
– Nick at Nite (50+)
–TV Land (50+)
34
Other Viacom Channels
• Target Audiences:
–BET (African American)
– Logo (Gay)
–Sundance (film fans)
35
8
I.1. Importance
and
Special Problems
of Demand
Estimation for
Media Industries
37
9
Why Demand Analysis?
• Every industry & firm wants
to know
–Who its potential buyers are
–What their willingness to
pay is
i
–What their price sensitivity
38
is
Why Demand Analysis?
–What product features they
value
–What they like about
competing
p
g pproducts
39
10
Why Demand Analysis?
- How to position its product
- How to plan the marketing and
promotion plan
- What the pricing strategy should
be
- Deploy its sales force
- How to select and manage
40
distribution channels
Why Demand Analysis?
- How to identify promotional
effectiveness
- How to identify market segments and
select target markets
• Etc
41
11
But it is Always Difficult To
Determine Demand
• It’s easy to graph a hypothetical
d
demand
d curve in
i a theoretical
th
ti l
economics model
• But very hard in the real world to
determine actual nature of
demand, and the factors that go
42
into it
“Assume a
Demand Curve”
Curve
P
Q
But Where Exactly Is It?
43
12
Demand analysis is
particularly important
( d diffi
(and
difficult)
l ) for
f media
di
and information firms
Wh ?
Why?
44
Recall the Fundamental Economic
Characteristics of Media
A. Supply Side
1.
High fixed costs, low marginal costs
2.
Convergent supply side
3
3.
Divergent cost in value chain
4.
Accelerating returns
5.
Excess supply
B. Demand Side
6.
Network effects
7.
Non-normal distribution of demand
C M
C.
Markets
k t
8.
Price deflation
9.
Intangibles
10. Public goods
11. Non-maximizers of profit
12. Role of government
Eli M. Noam, Mobility, 2006
45
13
1. High Investment Needs
and Uncertainty
• Media content is expensive
p
to
produce, is competitively
unique, and has short shelf
life.
–Demand estimation is
essential to reduce risk of a
project
46
Long Planning Horizons
• Presence of non-maximizers
of profit who will supply
products outside the market
• Continuous-flow products
(telecom services, cable TV,
newspapers, etc)
47
14
Long Planning Horizons
• require distribution networks,
strong economics of scale and
network effects andinvestment
far ahead of actual demand.
–Fiber-to-the-home
Fiber to the home
–Broadcast satellites
–Business plans
–IT equipment and semiconductors
48
Investment Uncertainty
• Outside investors must
evaluate projects (films, tech)
and companies by evaluating
the q
qualityy of the demand
forecasts.
49
15
http://realestatetomato.typepad.com/the_real_estate_tomato/80_20_principle.jpg
50
2. Instability of Preferences
1. Content suppliers must be
able
bl to
t rapidly
idl respondd to
t
changing audience tastes
51
16
3. Unique Products
• For each discrete-product media,
Product is unique
q
-Films, books, music
-Therefore separate marketing
“drives” necessary for each of
thousands of new products
• Many products are “intangibles” and
hard to evaluate in advance
52
g. Indirect Transactions
• “Public Good” characteristics
• Media
M di products
d t often
ft given
i
away rather then sold to
identifiable users. (e.g.,
b d ti )
broadcasting)
–Audiences must be identified
53
for advertisers
17
4. Unstable Markets
• “Excess supply”
pp y
• “Accelerating Returns”
• “Price Deflation”
• “Convergent
“C
Supply
S l
Industries”
–“Convergence of suppliers”
54
5. Technology Change
• For “new media”
and applications
–Rapid change
of technology
–Short
Sh t product
d t
cycles
http://www.rmh.de/media/intemplate/4_anim.jpg
55
18
Technology Change (Cont.)
• No consumer experience
with many new products
– e.g., MP3 players,
video cellphones,
p
, etc.
• Techno-optimism
(“push”)by producers
Iridium
56
6. The Subjective Value of
Information
• Information is an experienced
p
good.
g
Its value is only determined after
consumption.
g the value
• Thus, research revealing
of information prior to consumption
is important to media providers.
S. Rafaell and D.R. Raben. “Experimental Investigation of the Subjective Value of Information in Trading,”
in the Journal of the Association for Information Systems, Vol. 4, 2003, pp.119-139.
57
19
7. Supply Affects Demand
• Media create a buzz for their own
product and references shapes
audience
58
8. “Network Effect”
• Media demand is interdependent with that
of others:
– Telecom, Internet: benefits to users rise
with numbers of others on the network
–For Film, TV, Music, popular
M
Magazines
i
andd Books:
B k often
ft share
h
experience with peers; a major benefit
of media consumption is to be
connected with one’s peers.
59
20
Implications
• Leads to “extremes” of
success because of the way
users dynamically influence
each other.
De Vany and Walls, “Motion Picture Profit, The Stable Paretian
Hypothesis, and the Curse of the Superstar,” forthcoming in the Journal of 60
Economic Dynamics and Control, 2004.
“Network Effect”
• The average utility of the
service increases with the
number of other
participants. Therefore, the
demand increases with size
of networks. The more
people are on the network,
or share the experience, the
more people are willing to
pay.
Demand Curve
P
Q
61
21
• For these and other reasons,
demand analysis is
particularly
i
important
i
in
i the
media and information field.
• And particularity difficult
62
For more details see
Appendix A:
Special Problems in
Estimating Demand
63
22
64
23
B. Examples
p for the
Problems of
Forecasting Demand
66
Type I and
Type II Errors
67
24
“Type I Errors”: The
wrong action is
taken (accept
hypothesis
incorrectly)( A
“false positive”)
68
Type I and Type II errors
A Type I Error is the false rejection of a true null. It has a probability of alpha
(α). In other words, this error occurs as a result of the fact that we have to
somehow separate probable from improbable.
www.uwsp.edu/PSYCH/stat/10
25
• Most forecasts overestimate
the
h ddemand
d ffor products
d
rather than underestimate.
• Eternal optimism
Carey, John & Elton, Marin. “Forecasting demand for new consumer services:
challenges and alternatives.” New Infotainment Technologies in the Home. Demand70
Side Perspectives, Lawrence Erlbaum Associates, New Jersey, pp. 35-57.
Media Flops
71
26
Book Flops are Bi-partisan
72
Picture Phones
• AT&T (1963): “There will be
10 million ppicture phones
p
in
use by US households in
1980.”
http://research.microsof
t.com/users/jckrumm/i
mages/picturephone%2
0head.jpg
73
27
Satellite Phones
• The Wall Street
Journall (1998):
(1998) “The
“ h
consensus forecast by
media analysts is of
30 million
illi satellite
lli
phone subscribers by
2006.”
74
http://www.blueskynetwork.com/Images/Products/9500Pop.jpg
Estimating DVD Demand
http://images-eu.amazon.com/images/P/B0002VE5GW.02.LZZZZZZZ.jpg
75
28
Estimating DVD Demand
• In 2004, DreamWorks Animation
grossly over-estimated the DVD sales
for “Shrek
“
2.””
• Retailers returned millions of unsold
copies.
• DreamWorks fell short of earnings
forecasts by 25%
Merissa Marr, “How DreamWorks misjudged DVD sales of its monster
hit,” The Wall Street Journal, May 31, 2005 from Post-Gazette. 15, June
76
2005. http://www.post-gazette.com/pg/05151/513324.stm
29
“Type II Errors”: The
right
g action is not
taken
((reject
j hypothesis
yp
incorrectly)( A “false
negative”)
78
The Telephone
• Western Union,
world’s largest
g
telecom company:
there is no market for
the telephone. (1877)
79
http://www.fmd.duke.edu/images/contacts.jpg
30
Type I and Type II errors
A Type II Error is the false retention of a false null. It has a probability equal to
beta (β).
www.uwsp.edu/PSYCH/stat/10
Film
• Charlie Chaplin
(1916): “The
The cinema
is little more than a
fad. What audiences
really want to see is
flesh and blood on
the stage.”
81
http://www.doctormacro.com/Images/Chaplin,%20Charlie/Chaplin,%20Charlie%20(Gold%20Rush,%20The)_01.jpg
31
TV Invention
http://www.solarnavigator.net/inven
ntors/inventor_images/John_Logie_Baird_young_man.jpg
“For God’s sake go down to
reception and get rid of a
lunatic who
who’ss down there
there.
He says he’s got a machine
for seeing by wireless!
Watch him- he may have a
razor with him.”
-Editor of the Daily Express in
response to a visit by John Logie
Baird, 1925
82
TV vs. Film
• Movie mogul Daryl Zanuck
“[Television] won’t be able to hold on to
any market it captures after the first six
months. People will soon get tired of
staring at a plywood box every night.”
Darryl Zanuck
20th Century Fox studios
chief; 1946
83
Source: TIME, December 31, 1999
http://www.reep.org/resources/adv2001/images/angels/old_tv1.jpg
32
TV vs. Radio
• New York Times (1939): TV
will never compete with radio
since it requires families to
stare into a screen.
84
http://www.sfist.com/archives/images/old-TV-set.jpg
Computers
• “I think there is a
world market for
maybe
b fi
five
computers”
-Thomas Watson,
Chairman of IBM,
1943
Thomas Watson Library, Columbia
Business School
85
33
PC
• Ken Olsen, President, Digital
Equipment
q p
Corporation
p
((1977):
)
“There is no reason anyone
would want a computer in their
home”
86
Source: http://www.digidome.nl/images/Ken_Olsen-1.jpg
Source: http://ceee.gwu.edu/school_reform/kids_computer72dpi.jpg
Cell Phones
• McKinsey (1981) study for
AT&T: there will be only
900,000 cell phones in use worldwide by the year 2000.
• Reality: almost 1 billion
87
http://www.3g.co.uk/PR/April2003/Brick.jpg
34
PC
• “640
640 kilobytes of
memory should
be enough for
anybody ”
anybody.
- Bill Gates, 1992
http://derstandard.at/?url=/?id=1979631
88
Internet
• “Two years from
now, spam will
be solved.”
-Bill Gates,, 2004
http://derstandard.at/?url=/?id=1979631
89
35
90
Thus:
“Nobody
Knows
Anything”
(William Goldman,
Goldman
Hollywood Pundit,
1983)
91
36
True?
• Yes,, True
• But task is not to be exactly
right, but to reduce the
probability of Type I and
Type II errors
92
• To succeed against
g
competitors one need not be
always right
• Just a little less wrong
93
37
This Is The Subject Of
This Unit:
• How media and
communications firms can
improve assessing the demand
for their products and services.
94
95
38
But we must also keep
asking the question: should
media
di companies
i use
demand estimation
techniques like a car
techniques,
manufacturer or an airline?
96
• Shouldn’t media creations be
based on
– artistic judgment
– news judgment
– public responsibility
97
39
Critiques of Audience
Research
• Garrison Keillor:
“Guys in suits with charts”
have changed public
radio into an audiencedriven enterprise.
http://beyondwellbeing.com/al/garrison.keillor.gif
Alan G. Stavitsky, “Guys in Suits with Charts: Audience Research in U.S. Public Radio,” Journal of Broadcasting and98
Electronic Media, Spring 1995, pp/ 1-14
• Argues
g
that the focus on
audiences has ruined radio’s
“intellectual and moral
ggrowth,, ppassion,, variety,
y, and
pleasure.”
Stavitsky, Alan. “Guys in Suits with Charts: Audience Research in U.S. Public Radio.”
Aranet. Spring 1995. Journal of Broadcasting and Electronic Media. Last accessed 99
on 7
June 2007 at http://www.aranet.com/library/pdf/doc-0088.pdf.
40
• Doesn’t media create its own
demand, by influencing
people and their preferences?
Shouldn’tt it be ahead of the
• Shouldn
audience not following it?
100
• Is p
peoples’
p
demand shaping
p g
media content?
• Or is media content shaping
peoples’ demand?
peoples
101
41
• Social Science and
communications research have
not resolved this question.
• There is a continuous back-andforth between explanations
whether “powerful media” or
“powerful
powerful audiences”
audiences determine
media content.
Sonia M. Livingstone, “The Rise and Fall of Audience Research: An Old Story
With a New Ending,” Journal of Communication; Autumn 1993; 43, 4.
102
Entertainment as Play
• Psychological
y
g
Theory:
y Desire
for entertainment is an effect
of ancestral adaptations for
“pretend
p
pplay.”
y
Francis F. Steen, Stephanie A. Owens, “Evolution’s Pedagogy: An Adaptationist Model of Pretense
and Entertainment,” Journal of Cognition and Culture 1. 4 (2004): 289-321.
103
42
• Evolutionary psychology: desire
for “play” is an intrinsic human
character, because it is a crucial
feature and skill for human
survival.
http://www.stpeteha.org/images/Children%20pla
ying%20on%20sidewalk.jpg
Peter Vorderer, Christoph Klimmt, Ute Ritterfeld, “Enjoyment: At the Heart of
Media Entertainment,” Communication Theory 14:4, November 2004
104
• Entertainment is a form of
“pretend play,” allowing
people to gain experience that
they can use in future
challenging situations.
–Like
Lik a simulation
i l ti
Francis F. Steen, Stephanie A. Owens, “Evolution’s Pedagogy: An Adaptationist Model of Pretense
and Entertainment,” Journal of Cognition and Culture 1. 4 (2004): 289-321.
105
43
Predator Evasion
• Like in play-chase games,
where one functionally learns
strategic skills to evade or
defeat a predator or adversary
106
In Contrast, the Perspective of the
“Political Economy” and “Critical
Studies”
• Thee more
oe
political part of
communications
research (e.g.,
Frankfurt School)
b li
believes
in
i allll
powerful media
Max Horkheimer (L) and Theodor Adomo (R)
http://www.arikah.com/encyclopedia/Theodor_Adorno
107
44
The “Nielsen Approach”: the
powerful audience
• Audience ppreferences ggovern
• Media companies satisfy
these preferences
108
The Approach of “Cultural
Studies”: A synthesis
–Media “texts” are not passively
accepted by the audience.
–audience activity is involved in the
“encoding” process.
• The
Th meaning
i off media
di texts depends
d
d
on the cultural background of the
audience. (“Interpretive Communities”)
109
45
110
• For purposes of media
managment, both major
perspectives are correct
• Media
M di audiences
di
have
h
preferences
f
that can be analyzed
-This is called“ Media Research“
• But
B these
h
preferences
f
can also
l be
b
influnced
-This is called “Media Marketing“ 111
46
• This chapter
p deals with
“Media Research“
• Later, we will deal with
Media Marketing
112
113
47
The Late 1930s
• Studyy of modern
communications started.
• Became a new branch of
social sciences
Czitrom, Daniel. Media and the American Mind. Chapel Hill: University of North114
Carolina Press, 1983, p. 122-146.
Audience Preference Research
http://www.cba.unnl.edu/about/publications/emag/Volume2/Issue1/im
mages/ggallup.jpg
• The first
audience
di
studies were
performed by
George Gallup
when
h tteaching
hi
psychology in
Iowa.
Dennis, Wayne. Current Trends in Social Psychology. Pittsburgh: University of 115
Pittsburgh, 1948, p. 218-273.
48
Paul Lazarsfeld
• A central figure in
the development of
marketing
k i studies
di
in the 1930s.
• Emigrated to the
United States and
started an institute
at Columbia to
research radio.
http://www.fathom.com/feature/35683/1576_Lazersfeld_lg.jpg
Czitrom, Daniel. Media and the American Mind. Chapel Hill: University of North116
Carolina Press, 1983, p. 122-146.
117
49
For Details see
Appendix B:
Demand for Media:
Deeper Motivation
118
For Further Details see
Appendix H:
Behavioral Economics
119
50
120
How Media Companies
Organize their Demand
Research
121
51
Viacom’s Research Focus (from
it’s Annual Report)
• Audience acceptance
p
of pprograms
g
• Effectiveness of expenditures by
advertisers.
• Effectiveness of media co
co’ss own
promotion
Source: Viacom 2006 report
122
• Large
g media companies
p
engage in substantial
audience research at every
stepp
• [Details]
123
52
• Seven distinct types of research
11.
2.
3.
4.
5.
6.
7.
Concept testing
Positioning Studies
Focus group tests
Test screenings
Tracking surveys
Advertising testing
Exit surveys
124
Robert Marich, “Marketing to Moviegoers” Elsevier, “Distribution to Theaters”
125
53
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
IV. DEMAND EXPERIMENTS
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
126Need?
• Is This What Media Firms
II. Analytical
y
& Statistical
Models
127
54
And what do media
researchers do?
• We will now discuss a number
of techniques for analyzing
demand.
128
• Approaches
pp
range
g from
• a hands-on physiological/medical
• to abstract statistical, analytical,
model building technique
129
55
• On
O th
the one extreme,
t
PsychoP h
Physiology Testing
130
IV.5. Psychoy
Physiology Testing
131
56
Measuring the
audience’s
di
’
physiological
response to a
media experience.
132
A. Heart Rate (HR)
http://josephhall.org/images/bp_hrt.jpg
Niklas Ravaha, “Contributions of Psychophysiology to Media Research: Review and
133
Recommendations, ” MEDIA PSYCHOLOGY, Vol. 6 No. 2, 2004, pp. 193–235.
57
B. Electrodermal Activity
(EDA)
•
http://www.electrodermology.com/pics-new/biotronprobe-drop.jpg
• Skin conductance of electricity increases
when sweat increases due to arousal.
134
Electrodermal Activity (EDA)
• Measures responses to various
stimuli
i li (sudden
( dd noise,
i
emotionally charged visuals,
pain, anxiety, fear, guilt etc.)
http://www.bsu.edu/web/00t0holtgrav/317/physio.ppt#6
135
58
http://www.wearable.ethz.ch/education/sada/Emotion-Board
136
Electrodermal Activity (EDA)
http://web.axelero.hu/lavender/kpt/hallgatokhoz/vassy/weboldal/H7KLFI1.JPG
EDA measures of “before”, “during”, and “after” responses to an
137
emotional picture and a calm picture
59
Facial electromyography
(EMG)
• An electromyograph detects
the electrical potential
generated by muscle cells
when cells contract.
contract
138
http://www.Wikipedia.org
D. Respiratory sinus arrhythmia
irregularity
• Index of parasympathetic nervous
system (PNS),
(PNS) that can be related to
emotion.
139
http://www.biosvyaz.com/Htm_En/Sl_En/Sl02E03.gif
60
F. Electroencephalographic
(EEG) Activity
• Measures
brainwaves using
electrodes.
http://www.nexstim.com/images/prod_eeg_01.jpg
140
http://www.blackwellpublishing.com/abstract.asp?aid=161&iid=4&ref=0956-7976&vid=10
Electroencephalographic
(EEG) Activity
• Emotions can be observed by
frontal EEG activity
http://membres.lycos.fr/choppin/research/emotexprinterf.gif
141
http://www.blackwellpublishing.com/abstract.asp?aid=161&iid=4&ref=0956-7976&vid=10
61
• The first 3 of these
measures are easily
applicable and most
commonlyy used in media
research.
Niklas Ravaha, “Contributions of Psychophysiology to Media Research: Review and
142
Recommendations, ” MEDIA PSYCHOLOGY, Vol. 6 No. 2, 2004, pp. 193–235.
• Usually, no single
psychophysiological method
is enough. Often several
methods are used to identify
different responses.
Niklas Ravaha, “Contributions of Psychophysiology to Media Research: Review and
143
Recommendations, ” MEDIA PSYCHOLOGY, Vol. 6 No. 2, 2004, pp. 193–235.
62
• On
O th
the other
th extreme
t
from
f
these
th
physiological experimentation is
statistical model building
144
Analytical & Statistical
Models
A.
B.
C.
D.
Statistical Interference
Econometric Modelingg
Conjoint Analysis
Diffusion Models
145
63
146
II.1.
II
1 Statistical
Inference
147
64
Audience Research Methods
1930: Methods
d l d by
developed
b Paul
P l
Lazarsfeld, Bureau of
Applied Social
R
Research,
h Columbia
C l bi
University; and Frank
Stanton, CBSDied December 2006
Paul Lazarsfeld,
Columbia
148
Frank Stanton, CBS
Reasons for sampling instead
of doing a population census
–Cheaper
–Faster
Faster
–More practical
• But:
–Incomplete
Incomplete coverage
–Respondents could be
unrepresentative of
population
149
65
Population: The entire
group we are interested in
Example: US Households
Sample: Smaller group
selected for observation
Example: Nielsen panels
150
How Do We Get From a
Sample to an Estimate of the
Overall Population Parameter?
•Suppose
Suppose one takes 3 independent samples of the
same population.
•Question: Did you watch last week the “Golden
Age” Channel?
•But the samples may not be representative.
representative
Population: 300 Million people
Sample 2
5000 people
Sample 1
Sample 3
5000 people
5000 people
151
66
Sampling Statistics
• Sampling results would differ
slightly, “luck of the draw”
• But one would
ld expect that
h all
ll
three samples would yield a
similar estimate because drawn
from the same ppopulation
p
- Sample 1: p = 25%
- Sample 2: p = 27%
- Sample 3: p = 24%
152
Central Limit Theorem
n=22
n
n=10
n=55
n
States that the distribution of a variable found
in a sample approaches a “normal” distribution
as the number of samples increases
n=15
n=40
153
67
Case Discussion
• How many viewers tuned into
the “Golden Years Channel”
last week? The Nielsen panel
has 5000 households and
1250 of them sayy they
y
watched at least some of GYC
last month.
154
Percent Watching
GYC
p
pp̂ = sample
x
proportion
pˆ =
n = sample
n
size
x = positive
1250
response
pˆ =
= 0.25 or 25%
5000
155
68
But need to consider the
probability of a sampling error
p = pˆ ± e
•Where
pˆ : audience
di
share
h in
i sample
l
p: audience share in the population
e: margin of error
156
Sampling Error
•Sampling error (e)
–gives
gives us some idea of the
precision of our statistical
estimate.
157
69
Potential Error in Estimate
• (e) = potential error,
due to sample being
off
“off”
• z-score: indicates how
far an item is deviated
from its distribution
mean
•Population is • (p) = proportion that
large compared answered positively
to sample size • q=(1-p) those who 158
answered negatively
pq
e= z
n
• Only the sample size has any
effect on the margin of error
• The larger the sample size,
size the
smaller the potential for error
pq
e=z
n
i
Iff n increases
e decreases
159
70
Suppose these are the
parameters
p
p=.25
(25%
(
of sample
p watched))
q=.75 (75% did not watch)
n=5000 (sample size)
z=1.96 for a 95% probability
160
Case Discussion GYC…
e = 1.96
.25 × .75
=.012 or 1.2%
5000
161
71
Estimated Audience
• Assume 100 million HH in the
US,
S then
h the
h number
b off
American HHs that watched
GYC last month
–With
With 95% certainty
t i t
–Lies between 23.8 and 26.2
million (25 mil ± 1.2 mil) 162
163
72
For Details see
Appendix C:
Sampling
164
165
73
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
IV. DEMAND EXPERIMENTS
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
166Need?
• Is This What Media Firms
From such relatively simple
statistical
i i l tools
l with
i h a simple
i l
variable as a yes/no binary choice
were expanded to multivariable
analytical
l i l methods
h d
167
74
II.2.
E
Econometric
ti
Demand
Estimation
168
Econometrics is Estimation of
Statistical Relations of Several
Variables
• Method requires cross-section
over multiple data points or
time series analysis
169
75
• Synthesize large amounts of info
in an effective way
• provides framework for
systematic thought
– assumptions explicit
170
• Can use numerous variables
• Identify, track, and model key
variables (price,
(price competition,
competition
etc.) that affect demand, and
put them together in different
scenarios
171
76
172
Ordinary Least Squares (OLS)
• Use linear regression models
to quantify linear relationships
among variables
• Can estimate OLS regression
using
i statistical
t ti ti l software
ft
packages (STATA, SAS,
EXCEL, Minitab, etc.)
173
http://www.chass.utoronto.ca/~murdockj/eco310/F03_310_six.pdf
77
Typical Regression Analysis
Unit sales = a + b1 price + b2 advertising +
bi other variables + e
or
Market share = a + b1 lagged market share
+ b2 price + bi other variables + e
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide174
to
Profitable Decision Making,” Second Edition 1995
Other Control Variables
• Adding variables that might
affect sales,
sales such as
–Growth in GNP
–Growth in population
–Season
–Income level
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide175
to
Profitable Decision Making,” Second Edition 1995
78
• Demographic characteristics
include age, education, gender,
marital status
• Psychographic
P h
hi characteristics
h
i i are
concerned with the individual’s
lifestyle preferences- their
p
,
activities,, interests and opinions,
which marketers refer to as
consumer AIOs.
176
Logarithmic Models
Sales=
1
2
a (price) b(advertising) b (other variables)
Which is the equivalent of
ln sales =
ln a + b ln price + b advertising + b ln other + u
1
2
i
[ln is the “natural logarithm”]
177
79
• The coefficients of the
logarithmic models are the
elasticities (here of sales with
respect
p to pprice,, advertising
g
expenditures, etc.) and to
other variables
178
179
http://www.amosweb.com/images/ElDm33c.gif
80
Lots of Different Models For
Econometric Demand Estimation
• OLS
• Inverse
• Stone-Geary
• Quadratic
• Stochastic
• Discrete
• Dynamic
• Inter-temporal
•Engel
•Log-linear
•Semi-log
•Constant elasticityy
•2 stage least
square
•Etc., etc.
180
A. Estimation of
Demand Curves
Measuring Price Sensitivity
181
81
Example: Demand Estimation
for Newsprint (paper)
- For newspapers,
directories etc.
http://homepage.mac.com/albertkwa
n/Chronicle_Blog/C1258471436/E1
867671640/Media/newspaper%20ro
ll.gif
http://www.andrewdegrandpre.com/newspaper_roll_centered1.jpg
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
182
supported by Fisher Center for the Strategic Use of Information Technology.
• Of great importance to
newspaper companies:
- What will be the price of
newsprint paper?
• Also of great importance to paper
and forestry companies which
must make long-term
investments in new trees.
183
82
Approaches to Forecast
Newsprint Demand
1. The classical model: (FAO model)
(UN’s
’ Food & Agriculture
Organization) estimated demand for
newsprint as based on income levels
(GDP)
• Since GDP is rising, demand is also
rising
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
184
supported by Fisher Center for the Strategic Use of Information Technology.
Trends
• But in fact the newsprint
demand turned negative after
1987, despite rising GDP.
• So FAO model
did not predict
well
http://unadorned.org/morningpaper/images/papers/mp_200
30707_2.jpg
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
185
supported by Fisher Center for the Strategic Use of Information Technology.
83
A Second Model: the
“Regional Plan Association
(RPA) Model”
186
• “Print media price index” –
calculates the impact of changes
in print industry input prices,
which affects the printing and
publishing industries, and in
turn newsprint demand
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
187
supported by Fisher Center for the Strategic Use of Information Technology.
84
Here is how the two
models describe the past
and project the future
188
Figure 1. US Newsprint Consumption Projections: FAO (1995-2010 and
RPA (2001-2020)
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
189
supported by Fisher Center for the Strategic Use of Information Technology.
85
3rd Model Type
Newspaper Circulation Model
• Looks to newspaper
p p
circulation to explain changes
in the newsprint market.
• Since 1987,
1987 there has been a
decline in the volume of
newspaper circulation.
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
supported by Fisher Center for the Strategic Use of Information Technology.
rd
3
190
Model:
ln(d news ,t ) = γ 0 + γ 1Δ ln(circnews ,t ) + γ 2 ln(d news ,t −1 ) + μt
191
86
• A 1% increase in newspaper
p p
circulation would lead to a very
large increase (3.1%) in demand
for newsprint
p
Lauri Hetemäki & Michael Obersteiner, “US Newsprint Demand Forecasts to 2020,”
192
supported by Fisher Center for the Strategic Use of Information Technology.
Newspaper Circulation Model
• Several variants of the
Newspaper Circulation Model
(Models #4,8,9) explain
demand still better
193
87
Lauri Hetemäki & Michael Obersteiner, US Newsprint Demand
Forecasts to 2020, p.30.
194
195
88
Econometric
E
i Example
E
l
#2: Live Entertainment
196
197
http://i93.photobucket.com/albums/l60/stoy17/Ted/TedSaluteSlideSho.jpg
89
Demand for Live
Entertainment
• Model: Ui= f(Lei, OGi, zi)
• Ui is
i the
th utility
tilit off the
th person i
• LE is the “vector of live entertainment
purchased in the market.”
• OG is the “vector of other goods
purchased in the market.
market ”
• Z is the overall tastes pattern of the
people.
Cameron, Samuel. “Determinants of the Demand for Live Entertainments: some survey198
based evidence.” Economic Issues 11, no. 2 (2006): 51-64.
Demand for Live Entertainments
Cameron, Samuel. “Determinants of the Demand for Live Entertainments: some survey199
based evidence.” Economic Issues 11, no. 2 (2006): 51-64.
90
Demand for Live Entertainments
Cameron, Samuel. “Determinants of the Demand for Live Entertainments: some survey200
based evidence.” Economic Issues 11, no. 2 (2006): 51-64.
Demand for Live Entertainments
Dependent Variable = 1 If attend > 12 or more events per year; 0 otherwise. Estimation method: ML
Coefficient
Standard Error
LEEDS (dummy = 1 for Leeds)
Variable
-.940
1.405
TVHRS (hours of TV watched per week)
.036
.032
RADIOHRS (hours of radio watched per week)
-.009
009
.022
022
ALONE (dummy=1 if regularly attends events alone)
-.515
1.616
NUMPARTY (number of people in a party for an evening out)
.076
.108
URGE (maximum price would ever pay for a ticket divided by
-.005
.005
RSNPRICE)x100
RSNPRICE (idea of a reasonable price for a ticket for an evening out)
FEMALE (dummy=1 if female)
SINGLE (currently single)
-.172
.100
-17.915
7.928
1.658
1.355
GROSSINC (gross income of family unit)
.000
.000
NOCCUP (no current occupation)
-.611
1.300
DEGPLUS (highest qualification is a degree)
-.351
.875
AGE
-.272
.158
AGESQ
.003
.002
91
Demand for Live
Entertainment
• Findings:
g income effects were not
noticeable; going alone or in a large
party did not have an effect.
• Age did not have a significant effect
either
either.
• As people get older they may go to
less rock concerts but to more operas
Cameron, Samuel. “Determinants of the Demand for Live Entertainments: some survey202
based evidence.” Economic Issues 11, no. 2 (2006): 51-64.
Price
• The findings for
price were
interesting.
• The coefficient
for price was
negative for
males
l but
b positive
ii
for females.
http://new.krcgonline.com/uploadedImages/Shared/Shows/Price_Is_Right_Logo.jpg
Cameron, Samuel. “Determinants of the Demand for Live Entertainments: some survey203
based evidence.” Economic Issues 11, no. 2 (2006): 51-64.
92
204
Econometric
Example #3 What are
the Effects of General
Economy on
Advertising Volume?
205
93
206
• An econometric studyy of 8
major countries (Picard 2001)
finds that advertising
spending
p
g declines 5% for
each 1% reduction in GDP.
207
94
Effects of General Economy of
Advertising (cont.)
• Strongg correlation found for
Germany, Spain, Italy,
Finland
• Moderate correlations: UK
France
• Low correlation: Japan
208
• Print media most affected by
GDP
• 15% decline for 1% decline of
GDP, on average
–in
in US lower effect of GDP,
only 5.5% for newspapers,
2.5% for magazines
209
95
• Electronic media less affected
–4% TV (US, 3%)
–8% radio (US, 2.5%)
http://wifinetnews.com/images/reciva_net_radio.jpg
http://images.amazon.com/images/P/B00061ZNV
E.01.LZZZZZZZ.jpg
210
211
96
•Econometric
Example #4:
Competing Video
Games
212
Nintendo and Sega
• Assume both Nintendo and Sega
g
are competing in the home video
game industry. Either Nintendo’s
or Sega’s demand is determined
b bboth
by
th firms’
fi ’ currentt prices
i
andd
advertising expenditures.
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
97
The Demand Model
• In a situation of two competing
p
g
home video game firms, the
demand model for each firm is:
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
Variables of the Demand Model
•
•
•
•
•
•
Qit – firm i’s demand at time t;
Pit – firm ii’ss price at time t;
Ait – firm i’s advertising expenditures at time t.
α – parameter for brand-specific effects
η and β – own price and advertising elasticities
ε and γ – cross-price and cross-advertising
elasticities.
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
98
Parameters
• βit < 1, diminishingg marginal
g
returns to advertising
• γit < 1, diminishing marginal
returns to advertising
ηit > 1, εit > 1, own price elasticity
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
Price Sensitivity
• Sega’s
g pprice sensitivity
y is
relatively smaller than Nintendo’s,
because customers are more
willing to pay more for a product
t h l
technology
supported
t d by
b a large
l
network of users.
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
99
Advertising Effectiveness
• Similarly,
y Sega’s
g advertisingg is
also more effective compared to
that of Nintendo, because bigger
company can maintain its demand
with
ith less
l advertising
d ti i expenditures.
dit
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
Strategic Interaction
• This kind of competition
p
between
two firms contains strategic
interaction. So both firms may
want to actively manage and
l
leverage
its
it customer
t
base.
b
Shankar, Venkatesh and Barry L. Bayus. “Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry”, January 2002.
100
Example
E
l #5:
#5 Modeling
M d li
Film Box Office
220
• Studios estimate film’s revenues
based on previews, the
performance of previous movies
into the same genre, with the same
talent, similar characters, etc.
• Models based on life-cycle of
221
similar movies.
101
Computer Models for
Predictive Film Success
• Motion Picture
Intelligencer (MIP)
• MOVIEMOD
• Many others
http://www.adangio.com/galleryImg/large/movie175.jpg
Daniel B. Wood, “Can Computer Help Hollywood Pick Hits?” The
222
Christian Science Monitor, January 3, 1997, p.1
•
Tool to help
strategy
gy based
on the ticketbuying
behaviors of
past movies
http://www.nyjet.com/move%20tickets.jpg
Daniel B. Wood, “Can Computer Help Hollywood Pick Hits?” The
223
Christian Science Monitor, January 3, 1997, p.1
102
• Models to predict which movie
scripts will be hits and which will
be flops
“Revenge of the Nerds’ Part V: Can Computer Models Help Select Better Movie
Scripts?”Knowledge@Wharton, 29, November 2006. University of Pennsylvania
224
How do the models work?
• The methods behind the
models are proprietary and
unisclosed.
“'Revenge of the Nerds,' Part V: Can Computer Models Help Select Better Movie Scripts?”
Knowledge@Wharton. 29 November 2006. University of Pennsylvania.
225
103
Behavioral Representation of Consumer
Adoption Process in MOVIEMOD
Eliashberg, Jehoshua; Jonker, Jedid-jah; Sawhney, Mohanbir S.; Wierenga, Berend.
“MOVIEMOD: An Implementable Decision-Support System for Prerelease Market Evaluation
of Motion Pictures.” Marketing Science, Summer2000, Vol. 19 Issue 3, p226-243, 18p, 9 226
charts,
1 diagram; (AN 3623791)
• MIP tries to factor in
advertising expenditures,
number of theaters used in a
release, time of year of the
release, or competition from
other movies.
• Based on ticket-buying
behaviors for past movies.
227
104
MOVIEMOD
• Unlike other forecasting models for
films, MOVIEMOD needs no actual
sales
l data.
d
– But survey data from focus
groups
Mhanbir S. Sawhney, Jehoshua Eliashberg, Jedid-Jah Jonker, Berend Wierenga. “MOVIEMOD: An Implementable 228
Decision
Support
System
for Prerelease
Market
Erasmus
Universiteit
Rotterdam, December
1997 Evaluation of Motion Pictures” in Marketing Science. Vol. 19, No. 3, 2000
MOVIEMOD
• Subjects are exposed to
different sets of information
stimuli and are actually shown
the movie.
• They fill out post-movie
evaluations, including word-ofmouth intentions.
Mhanbir S. Sawhney, Jehoshua Eliashberg, Jedid-Jah Jonker, Berend Wierenga. “MOVIEMOD: An Implementable 229
Decision
Support System for Prerelease Market Evaluation of Motion Pictures” in Marketing Science. Vol. 19, No. 3, 2000
105
MOVIEMOD
• These measures are used to
estimate the word-of-mouth
word of mouth
parameters and other behavioral
factors, as well as the moviespecific parameters of the
model.
Mhanbir S. Sawhney, Jehoshua Eliashberg, Jedid-Jah Jonker, Berend Wierenga. “MOVIEMOD: An Implementable 230
Decision
Support System for Prerelease Market Evaluation of Motion Pictures” in Marketing Science. Vol. 19, No. 3, 2000
MOVIEMOD
• The heart of MOVIEMOD is an
interactive Markov chain model
describing the macro-flow process.
– allows to account for word-ofmouth spreaders in the population.
Eliashberg, Jehoshua; Jonker, Jedid-jah; Sawhney, Mohanbir S.; Wierenga, Berend.
“MOVIEMOD: An Implementable Decision-Support System for Prerelease Market
231
Evaluation of Motion Pictures.” Marketing Science, Summer2000, Vol. 19 Issue 3,
p226-243, 18p, 9 charts, 1 diagram; (AN 3623791)
106
Claims: The Dutch Application of
MOVIEMOD
• Managers used MOVIEMOD to identify
a final plan that resulted in an almost 50%
increase in the test movie’s revenue
performance
• The box-office sales resulted from the
final plan were within 5% of the
MOVIEMOD prediction
Eliashberg, Jehoshua; Jonker, Jedid-jah; Sawhney, Mohanbir S.; Wierenga, Berend.
“MOVIEMOD: An Implementable Decision-Support System for Prerelease Market
232
Evaluation of Motion Pictures.” Marketing Science, Summer2000, Vol. 19 Issue 3,
p226-243, 18p, 9 charts, 1 diagram; (AN 3623791)
233
107
Problems of Econometric
Demand Estimation
• Data
–Often insufficient
–Often
Often unreliable
234
• Need to assume a specific
mathematical model for the
relationship between price and sale.
• If specification is incorrect, the
results will be incorrect
• Predicting the future requires
assumption that behavior is like the
past.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide235
to
Profitable Decision Making,” Second Edition 1995
108
Problems of Econometric
Demand Estimation
• Econometric problems
–Serial correlation
–Multicollinearity
–Homoscedasticity
–lags
lags
–exogeneity
236
Problems of Econometric
Demand Estimation
• Results
–statistically significant?
–conclusion justified?
–Can one claim causality
–stable
stable over time,
time for
forecasting?
237
109
238
239
110
Case Discussion:
• How can Viacom use
econometric techniques to
estimate the demand for its
Golden Years Channel?
240
• A simple demand model could
be specified like this:
Likelihood of watching the Golden
Y
Years
Ch
Channel=
l
α + β1 ln age + β 2 ln income + β 3 ln education +
γ 1 adventure +γ 2 romance +γ 3 sports +
γ 4 documentaries/news + y1 pprimetime +
y2 daytime + y3late night + u +
e median age in zip code + f i other
241
111
• The coefficients that are estimated
are
βi = own-price elasticities to age, income,
education
δ= cross elasticity to other types of
channels
g
e = “network
“
effect”
f = effect of other factors z
u = error term
242
Y = time of day
• Some of the “other factors”
could be dummy variables for
yes/no of some factors, such
as “rural location,” “Latino” or
“living
living single
single.”
243
112
For Details see
Appendix D:
Econometric
Estimation
244
Measuring the Price
Elasticityy of Demand:
this is discussed in
detail in the Chapter on
“Pricing.”
245
113
246
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
IV. DEMAND EXPERIMENTS
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
247Need?
• Is This What Media Firms
114
II.3.
II
3 Conjoint
Analysis
248
Trade-off Analysis – Conjoint
Analysis
• Helps
p disaggregate
gg g a product
p
into the value given for each
attribute by consumers.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
249
115
• Developed initially by Paul
E Green and Vithala R.
E.
R Rao,
Rao
“Conjoint Measurement for
Quantifying Judgmental
Data ” Journal of Marketing
Data,
Research 8 (August 1971)
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
250
• Researcher asks respondent
to make choices between
different levels of two
product attributes.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
251
116
• Permits the researcher to
identify the value (utility)
that a consumer attaches to
each product attribute
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
252
• The value of a product is
equal to the sum of the utility
the
h consumers dderive
i from
f
all
ll
the attributes of the product.
253
P&B LLC DBA POPULUS http://www.populus.com/techpapers/conjoint.php
117
• This enables the researcher
to predict the prices which
the consumer would pay for
a product of various
combinations of attributes.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
254
• There are computer packages
(i.e. ACATM, Adaptive Conjoint
Analysis) that generate an
optimal set of trade-off
questions and interprets results.
255
118
Example #1: Attribute-Importance
Study For MP3 Player
(Scale 1-10)
Attribute:
Att
ib t
Quality:
8.24
Styling:
6.11
Price:
2.67
User Friendliness:
7.84
Battery Life:
4.20
Thomas
T. Nagle & Reed
K. Holden, “The Strategy and Tactics
of Pricing: A Guide256
to
Customer
Service:
5.66
Profitable Decision Making,” Second Edition 1995
Golden Media
• How could Viacom make use
of conjoint analysis for its
“Golden Years” channel?
257
119
Golden Media
• A cable company is considering
which package to offer to its
customers aged 65+. These vary
in:
– Price of package ($30-50)
– Movie
M i ffrequency (1
(1-4)
4)
– Golden Media channel (yes/no)
– Other channels (10-40)
258
Cable TV Package Options
Levels of attributes measured in survey
Attribute
Movie frequency
Level
1 per day
2 per day
Golden Age
channel
Yes
No
Price of package
$30
$14
$50
10 channels
20 channels
30 channels
Other channels
Source: According to P&B LLC DBA POPULUS
http://www.populus.com/techpapers/conjoint.php
3 per day
4 per day
40 channels
259
120
Conjoint Tasks
• Once data have been collected,
pparticipants
p
are given
g
to choose from
pairs of cable channels (conjoint
tasks).
• Each pprofile describes 2-4 attributes.
Participants are asked which of the
two channel descriptions they prefer
more.
260
P&B LLC DBA POPULUS http://www.populus.com/techpapers/conjoint.php
Computation of utilities
• Utilities are then calculated
by a statistical program.
Source: Kotler (1997), Marketing Management
261
121
Respondent’s utilities for
selected packages I
• For each package the overall
utility is calculated.
• Overall utility = Sum of all
weighted
i h d average utilities
ili i
262
Example: Cable TV Packages
Respondent’s utilities for selected packages II
Package Configuration
Nr. Other
channels
Utilities
Overall
Utility
Golden Movie aired Price
frequency
Age
channel
1
4 channels Yes
2 per day
$14
.471 + .769 + .271 + .035 =
1.546
2
4 channels No
3 per day
$12
.471 + .231 + .311 + .217 =
1.230
3
3 channels Yes
1 per day
$12
.403 + .769 + .103 + .217 =
1.492
4
3 channels No
4 per day
$12
.403
403 + .231
231 + .315
315 + .217
217 =
1 166
1.166
5
2 channels Yes
4 per day
$10
.125 + .769 + .315 + .738 =
1.947
6
2 channels No
3 per day
$10
.125 + .231 + .311 + .738 =
1.405
7
1 channel
2 per day
$10
.001 + .769 + .271 + .738 =
1.779
8
1 According
channel toNo
per day
$10
Source:
P&B LLC3DBA
POPULUS
http://www.populus.com/techpapers/conjoint.php
.001 + .231 + .311 + .738 =
1.281
263
Yes
122
• First ppackage
g would have
been the most attractive in
terms of content, but the price
is too high.
g
Source: According to P&B LLC DBA POPULUS
http://www.populus.com/techpapers/conjoint.php
264
• The configuration package number
5 with the lowest price, 20 extra
channels, the Golden Age channel,
and a movie frequency of 3 per day
is the most preferred,
preferred and most
likely to be chosen by the senior
consumer.
Source: According to P&B LLC DBA POPULUS
http://www.populus.com/techpapers/conjoint.php
265
123
266
For Further Details
see Appendix E:
Conjoint Analysis
267
124
268
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
IV. DEMAND EXPERIMENTS
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
269Need?
• Is This What Media Firms
125
II.4.
II
4 Diffusion
Diff i
Models
270
Generally, adoption of a new
product follows an S-Curve
Pattern
271
126
S-Curve Pattern of Adoption
• The S-Curve helps to illustrate and
to predict how a new product will
be accepted by the population
• The S-shaped curve of adoption
rises slowly at first when there are
few adopters
272
General Formula of the SCurve
a
Cumulative sales =
1 + be − kt
where t is time and a, b and k
are constants.
constants
McBurney, Peter, Parsons, Simon & Green, Jeremy. “Forecasting market demand for
new telecommunications services: an introduction.” Telematics and Informatics 19,
273no.
3 (2002): 225-249.
127
Viral Marketing operates on an
S-Curve
• Knowledge of the given thing will
spread like a “virus” epidemic
Wilson, Ralph. “The Six Simple Principles of Viral Marketing.” WilsonWeb. 1 February
2005. Last Accessed on 31 May 2007 at http://www.wilsonweb.com/wmt5/viral- 274
principles.htm.
• Also known as an “epidemic
p
model.” A “logistic” function
y(t) = N{1+0 exp [-kt]}
275
128
• Example:
p Adoption
p
of
Blue-Ray DVD
• Example: knowledge of a
hit movie
276
• With different parameters,
p
,
different S-shapes occur
• One has to determine, from
y data, what the pparameters
early
are, for a projection of the rest
of the S-curve.
277
129
Market Growth Curves
278
Problems
• finding acceleration point
• finding the “saturation level”
Carey, John & Elton, Marin. “Forecasting demand for new consumer services:
challenges and alternatives.” New Infotainment Technologies in the Home. Demand279
Side Perspectives, Lawrence Erlbaum Associates, New Jersey, pp. 35-57.
130
• comparison of the product to
b forecast
be
f
t with
ith some earlier
li
product that is believed to
have been similar
280
Example #1: DVD vs. VHS
• Can the diffusion of DVD
be compared to the
diffusion of VHS?
281
131
• VHS is in 95% of US HHs in
2008 (= Maximum Market
Demand);
–DVD penetration was 75%, in
2008
2008.
282
283
132
75 x 100
= 79%
–Thus, the HDI =
95
– Thus, the DVD market is
still 21% below its potential.
284
• VCR reached 75% after 12
years. DVD took only 6 years.
Hence DVD penetration rate is
2x faster than that of VCR.
• Since VCR took 3 years to rise
f
from
75% tto 95%
- hence, DVD is likely to take
only 4/2 =2 years to reach 90%285
133
• For Blu-rayy DVD,, can one
make similar comparisons to
DVD
• But,
But maybe consumers do not
value HD much over SD
quality?
286
• Problems with the diffusion
approach: There are too
many differentiating variables
to make comparisons
p
amongg
products have a strong
predictive value.
287
134
Case
Discussion:
289
http://www.bestchoicecare.com/library/images/tvcouple.jpg
135
Case Question: How would
“Golden Years” estimate and
measure its audience?
http://www.cdc.gov/communication/images/tv2.jpg
290
http://www.outsidein.co.uk/photos/sunray%20watching%20TV.jpg
Modeling the Market (I)
1. Identify Audience Age Cohorts
Million
People/
yr
Boomers
4 Mil
Classics
2 Mil
1 Mil
0
10
20
‘95
‘85
30
‘75
40
50
60
70
80
‘65
‘55
‘45
‘35
‘25
291
136
Modeling the Market (II)
2. Identify TV Viewing (Minutes/Day) By
Age
300
200
100
0
10
20
30
40
50
60
70
80
292
Modeling the Market (III)
3. Aggregate TV Hours by Cohort (# of
average TV hours/day x cohort)
Million
People
Aggregate TV Minutes/day By Age
1100
1000
450-
500
0
10
20
30
40
50
60
70
80
293
137
Advertisers Value Age Cohorts
Differently
• Younger audiences preferred
• Longer payback for investment
in customer acquisition
• Less
L rigid
i id consumption
ti routines,
ti
greater susceptibility to
advertising
294
Modeling the Market (IV)
4.
Value of TV Hours to Advertising by
Cohort (CPM x# of ads x# of hours)
Advertising Value of TV Audience by Age
Advertising minutes= 20% of TV minutes
Average CPM= 13$= 1.3¢/person/ad minute
CPM for 65: $8
CPM for 25-45: $16
Aggregate TV Min.
Average TV advertiser value of viewer/year= $200
Total TV advertiser value of US audience= $60 Bil/yr.
Total TV advertiser value of US Pop. 65+ years= ~ $4.2 Bil/yr.
Advertising Value of
Audience
0
10
20
30
40
50
60
295
70
80
138
• Each channel has a peak age
cohort A where it is viewed the
most. audiences declining at a
rate B away from the peak
cohort.
• The media firm can control A
and B through programming
decisions C is the size of the
decisions.
audience, and is a function of A,
B, and the presence of other
channels.
296
297
139
Audience, older & younger
(represented by the triangle)
T
V
M
i
n
u
t
e
s
C
B
Age
A
298
Modeling the Market (V)
5. Competitor Analysis
Aggregate TV Minutes for various Channels
by cohort (Schematic)
T
V
M
i
n
u
t
e
s
Cartoon
Nick Jr.
Fox
Nickelodeon
MTV
ABC
CBS
ESPN
CNN
0
10
20
Potential
advertising
value of
audiences
30
40
History
50
Age Cohort
60
70
80
299
140
Modeling the Market (VI)
6. Analysis of Under-Served
Niches
• Where are niches?
•Look for:
A. No domination by a strong brand
(e g Nickelodeon)
(e.g.
–Low peak of audience triangle (e.g.
History Channel)
B. Distance of competitors
300
from target cohort
Modeling the Market (VII)
7. Estimating market shares
• Make assumptions
- e.g. competitors that target
the same cohort share that
cohort equally.
• But that the share declines
with distance from the
target cohort
301
141
• Audience for a channel
depends on its positioning of
its peak at cohort i,
i with other
channels j in the market.
• For each cohort, its share is
d
determined
i d by
b the
h distance
di
off
that cohort from its peak
audience cohort
302
Model of Market Share:
% Share in cohort i by a channel =
S = ∑ TVi
i
100
∑ (1 − a ( PeakCohort
j
j
− Cohort i ))
j
303
142
• The channel’s
audience is
the
h sum off iits
share in each
cohort, times
the
h TV hours
h
of that cohort
http://www.awesomebackgrounds.com/templates/tv-channel-changer-01.JPG
304
i = age cohort
n = number of competitors
j = competitor j
b = coefficient of audience
specialization
(defines decline of % share by
distance of a channel’s peak
cohort)
(а can be measured for existing
channels; it is high for age- specific
channels, lower for inter305
generational channels (e.g., ESPN))
143
• Repeat this for every cohort i
• Total
T t l estimated
ti t d add revenues T
for channel:
Ti =∑ S %i x (# TV hours) i x CPMi
i
306
Management Decision Process
How to optimize Revenues T:
• Choose a combination of
–target peak audience cohort i,
– and the extent of audience
specialization (coefficient b)
» how steeply peaked the
audience triangle will be
307
144
• This model makes it possible
t check
to
h k outt multiple
lti l niches,
i h
and find the optimal niche,
and therefore the optimal
specialization
308
• The important
p
ppoint is to
think systematically and
break down the question of
channel strategy
gy into smaller
elements
309
145
310
• This is what analytical or
statistical modeling is about.
–Interprets data
• Good analysis requires good
data & its interpretations.
• This is the next topic: Getting
the Data
311
146
312
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
313Need?
• Is This What Media Firms
147
III. Empirical
Sampling of
Audience/
Consumers
314
III.1. Sampling
Methods
A.
B.
C.
D.
E.
F.
Personal Interviews G.
Mail and Phone
H.
Surveys
I.
Focus Groups
J.
Psycho-Physiology
T i
Testing
K.
Test Marketing
Internet Surveys
Retailer Surveys
Conjoint Analysis
Delphi Surveys
Trendsetters &
Opinion Leaders
Automatic
Metering
315
148
A. Personal Interviews
•In-home
316
http://www.ska-pr.com/personal%20interviews.htm
Mall Interviews
http://www.infonet.st-johns.nf.ca/providers/nhhp/newsletter/spring00/02_photo.gif
317
149
Major Players
• Personal surveys usually conducted
by market research firms, e.g.,
–Simmons
–Dun & Bradstreet
–Arbitron
–NFO
318
–Gallup
http://www.directionsmag.com/companies/images/logos/1252.jpg
319
150
http://www.wealthnationusa.com/xSites/Agents/wealthnationusa/Conte
nt/UploadedFiles/dun_and_bradstreet_logo.gif
http://www.dmwmedia.com/images/Arbitron.jpg
320
321
151
Personal Interviews Pro & Con
322
–
–
–
–
Can be indepth
Expensiveneed reliable
team
Sample
often biased,
selfselection
Follow-up
research is
timeconsuming
323
152
Problem with Personal Surveys
• The problem with most
surveys is that people will lie.
–about their income
–their
their taste
–Their actual viewing (or
they will be forgetful)
324
Mick Underwood, The Communication Studies Project, “Audience Measurement”
Other Problems With Personal
Surveys
“Interviewer effect”
-Age, gender, attractiveness,
pronunciation, intonation,
gestures etc.
- respondents
d
might
i h try to
impress the interviewer
325
Mick Underwood, The Communication Studies Project, “Audience Measurement”
153
• Futile to ask consumers what
they would be willing to pay for
a product.
product
• Direct questioning makes
consumers typically state a lower
price than they would actually
pay (bargaining
(b
i i behavior)
b h i )
–or, a higher price to please
interviewer
326
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide to Profitable Decision Making,” Second
Edition 1995
327
154
B. Mail and Phone
Surveys
(http://www.onesystem.com/)
328
Mail and Mailed Surveys
• Low-cost
• Greater anonymity increases
candor
• Low response rates lead to
bias
• For written surveys, no
probing or clarification
329
Mick Underwood, The Communication Studies Project, “Audience Measurement”
155
• Often used for new magazine
g
concepts, even before the
magazine is actually
ppublished in order to validate
concept and to get feedback
on price and features
330
Sample Test Mailing Grid
for Magazine
Mail pitch
Approach
Price
Offer
Content
A
A
Mailing Mailing
$10
$15
C
Mailing
$15
D
Mailing
$20
E
Mailing
$25
Soft
Soft
Hard
Soft
Soft
Broad
Narrow
Broad
Narrow
Broad
331
156
For more details
see Appendix
A
di F:
F
Direct Mail Test
Grid
332
Example #1: Telephone
Survey for Office Software
• A software
f
fi
firm developed
d l da
product for law firms that
would manage storage and
billi for
billing
f legal
l l documents
d
http://images.amazon.com/images/P/
B00005B0C6.01.LZZZZZZZ.jpg
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
333
157
• A random sample of 603
attorneys was contacted by
phone and asked for the
likelihood of purchase at either
$2000, $4000, $6000, or $8000
• About 150 responses per price
point.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
334
• The firm’s original intended price
was $500
• But survey showed that even at
$2000, 49% of the firms said
they would have bought the
package.
• Demand found to be highly
inelastic at high prices (see figure 335A)
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide to
Profitable Decision Making,” Second Edition 1995
158
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
336
• Price increase from $4000 to
$8000 did not change much
th proportion
the
ti off law
l firms
fi
willing to buy, but raised
sales revenue substantially
(Fi
(Figure
B).
B)
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
337
159
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
338
• Based on those surveyy
figures, what should the firm
charge?
339
160
Preliminary Conclusion:
• Charge
g $$8,000
,
• And also try to have a lowerquality product at about
$4 000
$4,000
340
• But problem: prices of
competing products are a
constraint
–can’t charge $8,000 if
competitor offers similar
product at $500
• Still the willingness to pay is
revealed
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide341
to
Profitable Decision Making,” Second Edition 1995
161
342
C. Focus Groups
• Recruited audience
g p makeupp is
- demographic
either random or selected
343
http://www.ctinfocus.com/images/foc.JPEG
Friedman, Motion Picture Marketing
162
Focus Group:
• Film previews
- 2 Types
¾Production previews: to help
managers and filmmakers
fine-tune the movie
¾Marketingg ppreviews: To
study audience’s reactions to
completed films, and assess
marketing strategy
344
Friedman, Motion Picture Marketing
Test Audiences
• Test Audiences are used byy
film companies to gauge
reactions to movies.
Bay, Willow. “What if ET died? Test audiences have profound effect on movies.” 345
Entertainment Weekly. 28 September 1998.
163
http://www.funworldmagazine.com/2003/Jun03/Features/Larger_Than_Life/images/A13Screen.gif
346
• Originally, Glenn
Close’s character in
“Fatal Attraction”
survived but
audiences hated her,
and the endingg was
changed to see her
die.
Bay, Willow. “What if ET died? Test audiences have profound effect on movies.” 347
Entertainment Weekly. 28 September 1998.
164
348
349
165
• Originally, ET died rather than
getting home in “ET”
• Originally Julia Roberts
dropped Richard Gere in
“Pretty Woman.”
Bay, Willow. “What if ET died? Test audiences have profound effect on movies.” 350
Entertainment Weekly. 28 September 1998.
Pretty Woman
166
Test Audiences Do Not
Always Prevail
• With “the
the Wizard of Oz
Oz”
test audiences complained
that “Somewhere Over the
Rainbow” slowed down
th movie.
the
i But
B t the
th song
stayed.
Bay, Willow. “What if ET died? Test audiences have profound effect on movies.” 352
Entertainment Weekly. 28 September 1998.
Director’s Perspective
“It’s much easier to embrace
the whole testing process
when you know that you
ultimately control the final cut
on your movies
movies. Buy it
it’ss
frightening if you’re in a
position where you’re going to
show the movie at a preview
and somebody else is going to
take the results of that
preview re-cut
re cut the film based
on that, maybe consulting you
or maybe not. That’s
terrifying.”
http://i.imdb.com/Photos/Events/4357/RonHoward_Grant_7604965_400.jpg
-Director
Ron Howard
Bay, Willow. “What if ET died? Test audiences have profound effect on movies.”
353
Entertainment Weekly. 28 September 1998.
167
National Research Group (NRG)
• NRG: film testing for Hollywood
distributors and producers
–Test screening of movies
–does most film testing
354
http://www.rxgetpaid.com/images/National-Research-Grouppaid-medical-research-logo.gif
355
168
Audience Perception
Analyzers
• These analyzers
y
are little,,
hand-held transmitters that
resemble TV remote controls.
Instead of buttons,, they
y have
a big dial on them.
Smith, Denise, “INSTANT ANALYSIS TECHNOLOGY HELPS RATE
COMMERCIALS” St. Louis Post – Dispatch, Jun 3, 1996. pg. 03
356
Audience Perception
Analyzers
• Linked to
software and
hardware that
registers the
responses and
their intensity.
Smith, Denise, “INSTANT ANALYSIS TECHNOLOGY HELPS RATE
COMMERCIALS” St. Louis Post – Dispatch, Jun 3, 1996. pg. 03
357
169
358
D. Using the Internet as
a Survey Tool
http://www.sphinxdevelopment.co.uk/Images/internetsurvey.jpg
359
170
Example: Nickelodeon
• Before production on a new version
of the TV series “Rugrats”
Rugrats began,
began
Viacom quizzed fans about what
they wanted
King, Tom, “Hollywood Journal: Nickelodeon Comes of Age --- At 20, Nick
Woos Big Stars, Takes On Old Studios; Building a Better 'Rugrat‘” Wall
360
Street Journal. Dec 1, 2000. pg. W.8
User-Level
Measurement
361
http://www.infosystem.gr/images/computer_user3.jpeg
171
The Data Meter
• In 1995, Media Metrix installed
the first meter of internet uses,
the “PC Meter,” into a consumer
sample
http://www.queensferry-pri.edin.sch.uk/nursery/photos/computer2.jpg
http://www.netprointer.com/image_file/seo_image/image021.gif
Steve Coffey, “Internet Audience Measurement: A Practitioner's View,” in
362
Journal of Interactive Advertising, Vol. 1:2, Spring 2001, p. 11.
• Requires user
cooperation.
• Incentives are
offered to users
who are willing
to use the
browser.
http://www.heart-disease-bypasssurgery.com/data/images/incentive.gif
Jaideep Srivastava, Robert Cooley, Mukund Deshpande, Pang-Ning Tan, “Web
Usage Mining: Discovery and Applications of Usage Patterns from Web Data” in 363
SIGKDD Explorations, Vol. 18: 2, January 2002 p. 13.
172
Major Tool: Cookies
• A standard programming device that
produces electronic files to tag individual
customers with a unique identification.
– Allows a website to recognize an
individual.
Deck, Cary A., “Tracking Customer Search to Price Discriminate.” Electronic Inquiry,
364
April 1, 2006.
Internet Surveys: Pro & Con
• Self-selection
• May require the respondent to
install special software.
James H. Watt & Michael Lynch. “Using the Internet for Audience and
Customer Research,” in T.J. Malkinson (Ed.), Communicating jazz:
365
p.127. New Orleans: IEEE.
173
Other Technique: Mouse
Activity Measurement
• Mouse Activities
- number of clicks
- time spent moving the
mouse in milliseconds
- time spent scrolling
http://www.dalveydepot.com/DalveyBMS.jpg
Mark Claypool, David Brown, Phong Le, and Makoto Waseda, “Inferring
User Interest,” in IEEE Internet Computing, Vol. 5:6, November 2001, p.
366
35.
367
174
Still Other
Types of
Surveys
368
E. Expert
E
Surveys: Comb
Analysis
l i
369
175
• “Comb Analysis”
- Technique for comparing
purchase criteria ((“most
most
important reasons for product
selection”) with opinion of
producer
Koch, Richard, The Financial Times Guide to Strategy. London: FT Prentice Hall, 2000,
p.193.
370
Comb Analysis
• E.g.
g If Dell wants to know
why it is selling fewer
computers to the Best Buy
retail chain than HP
http://www.sferaplus.hr/pr/hp/NotebookHPnc4000.png
371
176
Comb Analysis: 3 Steps
• First step,
p, researchers ask the
retailer to rate (e.g., on a 1-5
scale), the importance to its
customers of various
purchase criteria.
Koch, Richard, The Financial Times Guide to Strategy. London: FT Prentice Hall, 2000, p.54-7, 193.
372
Comb Analysis Example
Purchase Criterion Importance Score
Price
4.9
Strength of Brand
Name
4.5
Service
4.0
Product Innovation
36
3.6
Packaging
1.5
373
177
Comb Analysis:
1. Survey Retailers
5
4.9
4
4.5
4
3.6
3
Importance Score
2
1.5
1
Packaging
Product
Inovation
Service
Strength
of Brand
Name
0
Price
Importance Score
Purchase Criterion
374
Comb Analysis – 2nd Step
• Ask the pproducer ((Dell)) to
score the same criteria.
375
178
Comb Analysis
Purchase Criterion
Price
Strength of Brand
Name
Service
Product Inovation
Design
Importance
Score
Dell's Score
4.9
3.7
4.5
4
3.6
1.5
4.6
42
4.2
4
4
376
Comb Analysis
R t il Distributor's
Retail
Di t ib t ' Criteria
C it i and
d Dell's
D ll' Score
S
6
5
4
4.9
4.6
4.5
3.7
4.2
4
4
3.6
4
3
Retailer's
Assessment
Dell's Score
2
1.5
1
0
Price
Strength
of Brand
Name
Service
Product
Inovation
Design
377
179
• Dell seems to
over-invest in
design,
g , and
under-invest in
price cuts.
http://www.2shoptheworld.com/media/Dell-primoffer.jpg
378
Comb Analysis – 3rd Step
• Compare
p competing
p
g firms’
scores.
379
180
Comb Analysis
Purchase
Criterion
Dell’s
Score
Importance
Score
HP’s
Score
Pi
Price
49
4.9
37
3.7
5
Strength of
Brand Name
4.5
4.6
4.2
4
4.2
3.5
Product
Innovation
3.6
4
3.6
Design
1.5
4
2
Service
380
Comb Analysis: Competitor
Analysis
R t il Distributor's
Retail
Di t ib t ' Criteria
C it i and
d Dell's
D ll' v. HP's
HP'
Score
6
5
4
3
2
1
0
Importance Score
Dell's Score
Design
Product
Inovation
Service
Strength
of Brand
Name
Price
HP's
HP
s Score
381
181
Comb Analysis
• Comb Analysis indicates that
Dell needs to lower its price
(the most important purchase
criteria) to be competitive
with HP.
• But can cut cost of design
382
Comb Analysis
• If Dell lowers effort on
design (least important
purchase criteria), it could
lower pprice to Best Buy
y and
become more competitive
with HP.
383
182
384
F. Expert
p Surveys:
y
Delphi
385
183
Expert Surveys
Delphi Methodology
•Created in the
1950s by RAND
corp.
•Goal: Reach
expert consensus Apollo’s Temple in Delphi,
Home of the Greek Oracle
by experts on a
386
certain topic
Delphi Methodology
• Combines quantitative and
qualitative data
• Group process : 15 - 20
respondents
• Selected for their expertise and
experience
387
184
Delphi Methodology
• Anonymity of participants
• Written
i
responses to questions
i
• Direct communication between
respondents not allowed
388
Delphi Methodology
• First round of questions:
–Questions with answers of
scores 1-10
389
185
Delphi Methodology
• Second and subsequent rounds:
–Participants
Participants are provided with:
¾Information on how the entire group
rated the same item
¾Statistical feed-back related to their
own rating
i
¾Summation of comments made by
each participant
390
Delphi Methodology
•Given same questions again
•Delphi rounds continue until a
predetermined level of
consensus is reached or no new
information is gained
391
186
• The main benefit is that theyy
are quick and cheap.
• The negative is that they are
very highly speculative
speculative.
McBurney, Peter, Parsons, Simon & Green, Jeremy. “Forecasting market demand for
new telecommunications services: an introduction.” Telematics and Informatics 19,
392no.
3 (2002): 225-249.
But how good are expert
forecasts?
• Lord Kelvin, one of the world’s foremost physicists,
1895: “Heavier-than-air flying machines are impossible”
• Marechal
M
h lF
Foch,
h lleader
d off French
F
h military,
ili
1911:
1911
“Airplanes are interesting toys that are of no military
value”
393
Source: http://www.afa.org/magazine/graphics/0600korea8.jpg
187
• Astronomer Royal Richard Wooley:
1956: “Space travel is utter bilge”
Source: http://www.everett.wednet.edu/schools/high/everett/EHS_Files/STUDENT_WORK/moonwalk.GIF
394
• Lord Rutherford, Nobel
Prize Laureate: 1933:
“Anyone who expects a
source of ppower from
transformation of these
atoms is talking
moonshine”
395
Source: www.darvill.clara.net/nucrad/ images/rutherford.jpg
188
aJohn von Neumann, celebrated
scientist: 1956: “A few decades hence,
energy may be free, just like unmetered
air”
Source: www.ibm.com/ibm/history/exhibits/ chairmen/chairmen_4.html
Source: www.neuralmachines.com/ axon/signals.html
396
397
189
Case Discussion Golden
Years Channel: Delphi
Survey
398
• Need to select the Experts
p
– Gerontologists
– Marketers specializing in
retirees
– Social workers
399
190
Delphi Sample Questions
• “On a scale of 1-10, do retirees
get enough TV shows?”
shows?
• “Would they resent such shows
since it reminds them that they
are old?
old?”
• “How many hours a week would
they watch such shows on
average?”
400
401
191
G. Surveys of
T d tt andd
Trendsetters
Opinion Makers
402
SIGMA, “SIGMA, SIGMA Global Sensor” 31 Mar 2005, <http://www.sigma-online.com/en/
SIGMA_GlobalSensor>, 13 Apr 2006
403
192
Opinion Leadership
•Opinion leader is able to influence
others’
th ’ attitudes
ttit d or behaviors.
b h i
Source: M Solomon, Prentice Hall (1996),Consumer Behavior
404
Surveying Trendsetters
• Identifyy trendsetters ((ex:
celebrities, critics) and
determine their response
405
193
SIGMA, “SIGMA, SIGMA Global Sensor” 31 Mar 2005, <http://www.sigma-online.com/en/
SIGMA_GlobalSensor>, 13 Apr 2006
406
Trendsetters in the US
g
the
•“Affluent Progressives,”
“Emancipated Navigators,”
and the “Aspiring
Acquirers ”
Acquirers.
SIGMA, “SIGMA, SIGMA Global Sensor” 31 Mar 2005, <http://www.sigma-online.com/en/
SIGMA_GlobalSensor>, 13 Apr 2006
407
194
Trendsetting in Europe
• “In Europe, the members of the
U
Upper
Lib
Liberall Segment,
S
t the
th
Postmodern Segment and the
Progressive Modern Mainstream,
are responsible
ibl ffor mostt off th
the
trends.”
SIGMA, “SIGMA, SIGMA Global Sensor” 31 Mar 2005, <http://www.sigma-online.com/en/
SIGMA_GlobalSensor>, 13 Apr 2006
408
Trendsetting in Japan
• In Japan, the „Modern Rich“, the
„New Citizens “ and „Young
Urbanites “ are usually the origin
of trends.
SIGMA, “SIGMA, SIGMA Global Sensor” 31 Mar 2005, <http://www.sigma-online.com/en/
SIGMA_GlobalSensor>, 13 Apr 2006
409
195
Technorati.com
• Rates blogs.
• Ranks blogs
based on the
number of sites
that link to it.
http://www.customersarealways.com/uploads/technorati-thumb.gif
“How Does Technorati Work.” Last accessed on 18 June 2007 at
http://trailblogging.com/2007/03/28/how-does-technorati-work/.
410
Critics
Two alternative perspectives on
the role of critics.
1 Critics
1.
C iti could
ld be
b opinion
i i leaders
l d
who influence audience demand.
2. Critics could be predictors of
their respective
p
audiences.
-Critics wired to act more as
leading indicators than as
opinion leaders.
411
Jehoshua Eliasberg; Steven M. Shugan, Film Critics: Influencers or Predictors
Journal of Marketing (Apr 1997)
196
Research Study Findings:
• The % of positive and negative
critics reviews is a statistically
insignificant predictor of box office
performance for the early
(
1-4).
)
weeks(weeks
412
Findings:
• It is, however, a statistically
significant predictor of box office
performance for later weeks, and
for cumulative box office.
413
197
• These findings do not support
the “opinion leader”
perspective, which would
predict that the greatest
influence of the review should
be immediately following the
review.
• But it does support the
M
414
“predictor” hypothesis
415
198
H. Automatic
H
A t
ti
Audience
Metering
416
Audience Research
Purpose:
• To let broadcasters know who their
audience is, and how it responds
• To let broadcasters know hoe much
to charge for advertising
• To let advertisers know who they
are reaching
417
199
Lots of Money at Stake
Major TV Advertisers (2006)
•
•
•
•
•
•
•
•
•
•
Procter &Gamble
G
General
lM
Motors
t
Time Warner
Verizon
AT&T
Ford
Disney
Johnson & Johnson
DaimlerChrysler
GlaxoSmithKline
Source: Schiekofer, The Media Marketplace. New York: Mediacom
$4.6
$4 4
$4.4
$3.5
$2.5
$2.5
$$2.4
$2.3
$2.3
$2.2
$2.2
418
Paul F. Lazarsfeld
• Applied mathematician from
A
Austria.
i
• Central figure in the growth of
empirical social science.
• Integrated market research
with psychological analysis.
Daniel Czitrom. “The Rise of Empirical Media Study: Communications Research as
Behavioral Science, 1930-1960.” In Media and the American Mind. Chapel Hill, NC:
UNC Press, 1982.
419
200
Early TV audiences: Diary
System
• Traditional Nielsen
methodology especially for
methodology,
local TV markets.
- used 4x a year during
“sweeps”
sweeps periods for local
stations.
• viewers record TV viewing 420
1. Diary System
• opportunity for samples to lie
• misses responses from
children, travelers, and
TV viewing in bars
• difficult with channel surfing
421
201
Sample Bias
• In the past, response rates of 70%
for diaries.
• Today
T d it
i is
i difficult
diffi l to get 50%
response rate for a meter panel,
25% for a diary
• If the p
people
p who do not respond
p
view TV differently from those
that do, then the ratings are biased
and wrong.
422
2. Also used for TV “overnight”
ratings: Telephone Surveys
–Fast
–sample biased
–Respondents
run out of
patience
423
202
History: Dynascope
• 1965, the “passive audience
meter” called the Dynascope:
meter
a movie camera that took
pictures, of both the TV viewer
and the TV show everyy 15
seconds.
Erik Larson, “Watching Americans Watch TV,” The Atlantic Monthly,
424
March 1992
History: Dynascope
• 1.5 million pictures were
analyzed:
- When the TV was on, 19% of
the time no one was in the room.
- 21%
% off the
h time
i the
h person was
engaged in a different activity
Erik Larson, “Watching Americans Watch TV,” The Atlantic Monthly,
425
March 1992
203
History: Infrared
Scanners
• Kiewit
Kiewit’ss “hot
hot bodies”
bodies
- scanned for people with an
infrared sensor.
- But Kiewit’s scanner
distorted by the “big-dog
Erik Larson, “Watching Americans Watch TV,” The Atlantic Monthly,
426
effect.”
March
1992
http://homepage3.nifty.com/shibadog/Album2/Album32/wanloaf3.jpg
427
204
More Practical Solution:
The Nielsen People Meter
• pplaced on each TV set in a
sample household.
• an electronic system placed
in 5,000
5 000 randomly selected
households in the U.S.
428
People Meters: Pro
• instant
i t t measures
• no “lying”
429
205
People Meters: Con
• children, travelers, and bar
viewing not captured
• nobody may be watching
• requires viewers to identify
themselves
http://www.printphoto.com/contest_pics/finalist
0902/I'm%20Not%20Tired.jpg
430
431
206
People Meters: Con
• Older p
people
p have higher
g
refusal rate to participate
• Young men most willing to
employ meter
432
Also, the greater audience
fragmentation creates greater
relative unreliabilityy of results
• The % of standard deviation
tends to grow as ratings
become smaller
smaller.
433
207
• E.g.: a “true” ratings of 6, in
sample of 3,000, will show as
sample ratings between 5.2
5 2 and
6.8 (± .8) in 95% of samples.
–i.e. relative error is ±14%
• But same error for
f “true”
“
” rating
i
of only 2 (± .5) will have a
relative error of ±25%
434
• And for a small cable channel
with “true” rating of .3, ±.2, the
relative error is ±65%
http://www.webspin-design.com/assets/Newsletter/Sept03/nr-reach-trend-top.gif
435
208
Case Discussion:
People Make
for “GYC”
• In theory GYC could benefit from the fast and
relatively accurate TV ratings data via the
People Meter.
– would also show demographics
• In practice,
i its
i ratings
i
will
ill be
b too low
l to
register
436
Can ratings be
manipulated?
437
209
Japanese Rating Scandal
• In 2003 a producer of the
Nippon TV Network (NTV)
manipulated television ratings
for his show
“Heads Roll in NTV Ratings Scandal.” Japan Times Online. 19 November 2003. Last
438
accessed on 19 June 2007 at http://search.japantimes.co.jp/cgi-bin/nn20031119b6.html.
Japanese Rating Scandal
• The producer used money to find
out what specific household were
being observed by the ratings
agency Video Research Ltd. and
ggot those homes to watch certain
shows by bribing the occupants
through various benefits.
“Heads Roll in NTV Ratings Scandal.” Japan Times Online. 19 November 2003. Last
439
accessed on 19 June 2007 at htt p://search.japantimes.co.jp/cgi-bin/nn20031119b6.html.
210
3. Automated
Metering
• The first mechanical
device to measure TV
demand was the
Audimeter where a
Audimeter,
stylus scratched out a
record of radio tuning
http://www.desmoinesbroadcasting.com/xtras/nielsen-audimeter-fullpix.jpg
Erik Larson, “Watching Americans Watch TV,” The Atlantic Monthly,
440
March 1992
• The chairman of Nippon
pp
Television Network (NTV)
Corporation was forced to
resign
g
441
211
Broadcast Data System
(BDS)
• Used for the
Billboard Top
100 Singles
g
• Tracks songs
played on the
radio
http://www.covenantdesigns.com/marketing/top_100_9surf.jpg
Poltrack, David. “Media Audience Research” Course. Columbia University Business
442
School. Fall 1998.
Broadcast Data System
(BDS)
• The BDS is still used today as the
“Nielsen BDS” and tracks over
1,000,000 songs each year.
• Radio/artist managers request
over 10
10,000
000 reports each day.
day
• Some songs are big on radio but
not in sales.
“About Nielsen BDS.” BDSonline.com. Last accessed on 15 June 2007 at
http://www.bdsonline.com/about.html.
443
212
444
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
• Is This What Media Firms Need?
445
213
III.2. NewGeneration People
Meter: The Digital
Meter System
446
• Identifies audio and TV content
through active codes embedded
in the program itself and in the
commercial messages
• Search engines identify the
programs and the
advertisements that are
watched
447
214
• This enables real time reports
on watching or listening
• can meter broadcast,
broadcast DBS
DBS,
PVR, digital cable, and radio
use.
448
http://nbc.com/Friends/index.html
The Battle of the Meters
• Nielsen Local People Meter
(
(LPM)
) vs. A
Arbitron
bi
Passive
i
People Meter (PPM)
• Channel-based vs.
programbased
449
Source:ppm.arbitron.com
215
Nielsen LPM Procedure
• A meter rests on top of every
TV in a Nielsen household and
each family member has an
assigned number.
John Maynard, “Local People Meters May Mean Sweeping Changes on
450
TV,” The Washington Post, April 28, 2005, A01.
http://www.nielsenadvertiserservices.com/images/box_4.gif
• Old local station system diaries
collected in “sweep” periods
• Nielsen initiates overnight
Local People Meter data
–Larger
Larger local samples
(8000 vs. 540 for diaries)
451
216
Nielsen Local People Meter
(LPM)
•$30 mil development
•Permits collection of
audience response in near
real time.
•Continuous measurements of
major local markets (not just
for 4 sweep
s eep periods)
•Includes demographics
•Launched in Boston, 2002
•Full-scale operation in 2006
http://www.nielsenmedia.com/lpm/images/people%
20meter-new.jpg
452
• Includes low resolution optical
meter that monitors how many
people
p
p are in the room,, and
identification of members of
households
• Can determine fast
fast-forwarding
forwarding
through ads.
453
217
•Expanded national sample
from 5,000 to 10,000.
454
The Arbitron Portable People
Meter(PPM)
•Portable People Meter, is
worn by consumer, detects
and records programming
wherever consumer located
•And whatever the program
source
455
Source:ppm.arbitron.com
218
Arbitron PPM Page 513
http://digital-lifestyles.info/copy_images/arbitron_2-lg.jpg
456
PPM
• Portable people meter (PPM) tested
in Houston, in 2005/2006
• The
Th PPM reads
d an encoded
d d audio
di
message that is embedded into the
audio track of every piece of media
(including, for example, TV, radio
and the Internet) that has sound.
sound
Besser, Charles N., PPM is the next big score for sports TV. Advertising Age, Vol.
457
76 Issue 26, p22-22, 6/27/2005. VOD,
219
The Portable People Meter
System in Action
458
Source:ppm.arbitron.com
• Arbitron PPM (worn by users) is
better able to keep up with
–Multiple TV sets in household
–Out-of home viewing
• But requires uses to wear the
y
device or have it nearby
• more expensive, but can be used
for radio, TV, Cable, and others.
459
Source: Broadcasting & Cable, 2/2002
220
460
III.3. Metering
Alternatives: Cable
Box and TiVo Box
461
221
• Alternative: use the digital settopbox (STB) of cable or satellite
TV
• Would increase sample size to
hundreds of thousands per
market
• Concept and technology
introduced in 1980s (CUBE
cable system) in Columbus, Ohio
462
Set Top Box
http://www.comcast.com/MediaLibrary/1/1/About/PressRoom/Images/
LogoAndMediaLibrary/Photography/DCT700DigitalCableBox2.jpg
463
222
• CUBE data used in litigation and courts.
– Columbus, Ohio pornography trial:
“Captain Lust” was shown to be one of the
most popular programs
– New Haven, CT: Least watched “You and
the Economy” (A Panel of Yale economics
professors was watched by 3 HHs)
• Cable industry
ind str decided nott to collect STB
data, individually or in aggregate, to avoid
giving customers a feeling they are being
watched and monitored.
464
Most Popular Program in
Columbus, Ohio
465
http://www.moviegoods.com/Assets/product_images/1010/213997.1010.A.jpg
223
• First trial STB of multichannel real-time metering,
g,
1997 Atlanta
466
• The media research agencies
utilize aggregated set top box
data which it acquires from cable
operators to provide a second by
second-by-second
second
by second analysis of
viewing habits.
“MTV Networks Leverages Charter Data from TNS Media Research”,
Wireless News, August 10, 2007
467
224
• Shift from program ratings to
commercial ratings. Commercial
ratings is the ability to measure
how many viewers were tuned
when the commercial was
actually running.
running
George Shabbab, “Not A Second to Lose,” MediaWeek, New York: July 23- July
468
30, 2007
TiVo Box
• Enables real-time monitoringg
and historical data for a
month
• Permits analyzing of time
shifting and zapping of
commercial ads
469
225
TiVo Box
http://www.nytimes.com/images/blogs/tvdecoder/posts/1107/tivo-box.jpg
470
DVR Page 526
http://www.timewarnercable.com/MediaLibrary/4/55/Content%20Manag
ement/Products%20And%20Services/imagesDVR/dvr-mainbanner.jpg
471
226
Cellphone Use for Media
Measurement
• Usingg specially
p
y adapted
p cell
phones to measure what
consumers listen to and see
– Provider: Integrated Media
Measurement Inc
Clark, Don, “Ad Measurement is Going High Tech.” Wall Street
Journal, Section B; Page 2, Column 3, April 6, 2006, Thursday.
472
Real time viewingg
measurement for TV
programs
227
• Nielsen has also launched a new data
service Nielsen DigitalPlus which
integrates set top box data from cable
and satellite operators with TV
measurement data from Nielsen
Media Research, commercial activity
data from Nielsen Monitor Plus,
Retail and scanning information from
AC Nielsen and modeling and
forecasting information from Claritas,
Spectra and Bases.
Katy Bachman, “Nielsen to Roll Out DigitalPlus”, Mediaweek.com, February 474
12, 2007
Media consumption tracking:
Nielsen’s plans
• Nielsen intends to track consumers’
activities on the web, TV, mobile
and per GPS when shopping.
• They work with Ball State
University to observe people in
their homes.
Story, Louise. Nielsen Looks Beyond TV, and Hits Roadblocks. The New York Times, February 26, 2008
228
Media consumption tracking:
Nielsen’s plans
• Nielsen acquired
q
firm to track
people’s eye movements, brain
waves and perspiration, which can
be used for TV and internet
activity tracking.
Story, Louise. Nielsen Looks Beyond TV, and Hits Roadblocks. The New York Times, February 26, 2008
Media consumption
tracking: Limits
• An alternative from ggatheringg data
across all media from the same
consumers (demanded by customers
but facing resistance from
consumers)) is
i merging
i data
d t from
f
separate panels resulting in quality
loss.
Story, Louise. Nielsen Looks Beyond TV, and Hits Roadblocks. The New York Times, February 26, 2008
229
Media consumption
tracking: Limits
• Not to lose the established panel
p
participants, Nielsen has to balance
their thirst for data with their
understanding and respect for
consumers’’ privacy.
i
• The ideal of tracking consumers
across all media remains a dream.
Story, Louise. Nielsen Looks Beyond TV, and Hits Roadblocks. The New York Times, February 26, 2008
479
230
Measurement
Technology Affects
Results. Therefore, it
i a Battlefield
is
B ttl fi ld
480
Important Consideration
• Meteringg is not about
technology, but about money
• Any change in metering
procedure has economic
effects
481
231
Measurement Technology
Affects Results. Therefore, it
is a Battlefield
•Broadcasters vs. cable channel vs.
advertisers
•Nielsen in the middle
•For example,
p , the effect of the
adoption of the People Meter, over
paper diaries, was significant
•And the shift to LPM does the same
482
http://gr.bolt.com/oldsite/download/pc/action/battlefield_1942.jpg
Changes in Ratings Patterns
for Prime Time Before,
Duringg and After
f the
Introduction of the People
Meter
William Adams, Journal of
Media Economics, 7(2) 1328, 1994
483
232
Overall Effect of People
Meters on Ratings
• Permanently lowered overall
TV ratings in 1990 by an
average of about 4.5 points.
• CBS: lost 2.0
2 0 points:
NBC: showed avg. loss of 1.5
484
ABC: little effect
CBS Lost 2.0 Points in change to
people meter
http://i.afterdawn.com/v3/news/cbs_logo.jpg
485
233
NBC Lost 1.5 Points
http://www.midnightchimesproductions.com/MCP/images/NBC-logo.gif
486
Effects on Programming
Categories
• Participation shows were
boosted 5 points in rating;
sitcoms 1.5; news 0.2:
• All other categories
dropped. Medical shows
showed highest drop; -4.1
487
234
Business Impact
• In 1990, each ratings point was
worth approximately $140
million/yr
illi /
• Decrease in ratings could cost
major networks between $400
and $500 million/yr.
• Cable: ratings gain of almost
20%.
488
• Cable networks fear contentspecific ratings less than TV
networks because they are not
as dependent
p
on advertising.
g
Lowry, Brian, “The Ratings: Inside and Out; Analysis: Networks seem to have decided
489
the ratings battle wasn’t worth the effort,” Los Angeles Times, July 12, 1997.
235
*
Impact of Local People Meters
• Here, too new metering has major
impacts on numbers
• In NYC, Fox 5, UPN 9 and WB
11 showed big drops.
490
• For Washington D.C., the
claimed undercount rates were
25% for Hispanic homes and
20% for African American
homes.
John Maynard, “Nielsen Delays Release of Local People Meters,”
491
Washington Post, Thursday, June 2, 2005, C07.
236
• Washington D.C. 2005 tryout
(600HH) showed not 650,000
HH watched local TV from 55
7PM, but only 526,000.
• Cable lost another 114,000
HH
HH.
492
LPM Effects
• Fox TV network and several local
stations complained that LPM
undercounts
d
t minority
i it viewers
i
in
i
cities.
• Don’t Count Us Out, a group
f d d by
funded
b News
N
C
Corp., generated
d
political pressures in Washingtong
John
Maynard,
“Local People
Meters May Mean Sweeping Changes on
and
NYC
on Nielsen.
493
http://images.zap2it.com/2
0031016/fox_logo_240_00
1.jpg
TV,” The Washington Post, April 28, 2005, A01.
237
• To mollify its critics Nielsen
agreed to a R&D fund to
improve its methodology.
• Creation of an Advisoryy
Council
Katy Bachman, “Nielsen Outlines Changes to Ratings Service,”
Mediaweek, February 21, 2005.
494
495
238
• Thus one can see that ratings
technology
gy and ratings
g
methodology affect dollars,
Euros, and Yens
• It is therefore important that
the ratings agencies are
trusted by all sides
496
• Minimum standards for
broadcast audience analysis
research have been established
by the Electronic Media
Ratings Council in New York,
which audits and accredits
rating
i services
i
497
239
• Members:
–National Association of
Broadcasters
–Cable Advertising Bureau
–Television Advertising
Bureau
–Magazine Publishers of
America
498
For more details see
A
Appendix
di G:
G
Audience
M
Measurement
Firms
Fi
499
240
500
We’ve looked at how to
measure audiences.
audiences
Next question is, how
to interpret and use
the data
501
241
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
502Need?
• Is This What Media Firms
III.4.
III
4 Audience
Metrics
503
242
10 Audience Metrics
1.HUT
2.Rating
3.Share
4.GRP
5.CUMS
6. AQH
7. AF
8. CPM
9. Quads
10. Q
504
• Households are usually the base
unit not people when measuring
unit,
audiences.
• Audience measures are usually
done in parts of days.
• TV ratingg services ((ex: Nielsen)) set
their
h i own geographic
hi rating
i areas.
505
Television Ratings Lab. “Television Ratings.”
243
Important Television Ratings
Terms and Facts
• Ratings = (100 x Households
viewing
i i program)) divided
di id d by
b
(total households with TVs)
• Share of Audience = (100 x
households viewing program)
divided by (households using
TVs that instant)
506
Television Ratings Lab. “Television Ratings.”
#1-3:HUT, Ratings,
Shares
507
Nielsen Media Research
244
Audience Metric #2
1.
Viewers of a program
TV HH
• In US ~105 mil TV HH
• Example:
–20
20 mil HH watch E.R.
ER
Rating = 20× 100 =19.0
105
508
2. “Share” (of Audience)
• The percent of TV sets in use (or
persons viewing) tuned to a
program.
SHARE =
Viewers x 100
HUT
–HUT: Households Using TV
actually watching at that time509
245
Audience Metric #1: HUT
(Households using TV)
•Number of share
• example:
l 60 mil
il HH watchh any
TV during CSI time slot.
(=HUT)
–Share
Share = 20 mil HH x 100/60
mil HH (HUT) = 33.3
• Share > Rating
510
–since HUT < TV HH
Broadcast TV: Nielsen Media
Research Top 10
(Week of May 12, 2008)
Rank
1
2
3
4
5
Program
American
i
Idold l
Wednesday
American Idol-Tuesday
Dancing With The Stars
CSI
Dancing W/ Stars Results
Network
FOX
Rating
14.6
FOX
ABC
CBS
ABC
14.4
11.9
11.2
11.1
*Measured in millions; includes all persons over the age of two.
http://www.nielsen.com/media/toptens_television.html
246
Nielsen Media Research Top 10 (Week of May 12, 2008)
Rank Program
p
6
Desperate
Housewives
Grey’s Anatomy7
Thu 9PM
8
Without a Trace
9
NCIS
10
CIS: Miami
Network Rating
ABC
10.7
10
House-Mon 9PM
ABC
10.5
CBS
CBS
CBS
9.6
9.5
9.1
FOX
9.1
*Measured in millions; includes all persons over the age of two.
http://www.nielsen.com/media/toptens_television.html
Highest Ranked Regular
Program Series, US
1950 51
1950-51
1951-52
1952-53
1953-54
Texaco Star Theatre
Arthur Godfrey’s Talent Scouts
I Love Lucy
I Love Lucy
1991-92
1992-93
1993-94
1994-95
1995-96
1996-97
60 Minutes
60 Minutes
Home Improvement
Seinfeld
E.R.
E.R.
Share
61 6
61.6
53.8
67.3
58.8
21.7
21.6
21.9
20.4
22.0
21.2
Rating
81
78
68
67
36
35
33
31
36
35 513
247
Highest Rated Individual
Broadcast
Top 10 US Telecasts 1960-1990
Rating
R
i
60.2
53.3
51.1
49.1
48.6
48.5
48.3
47.7
47.4
47.2
1 MASH Special
2 Dallas
3 Roots, PT VIII
4 Super Bowl XVI
5 Super Bowl XVI
6 XIII Winter Olympics
7 Super Bowl XX
8 Gone With The Wind, Pt. 1
9 Gone With The Wind, Pt. 2
10Super Bowl XII
Share
Sh
77
76
71
73
69
64
70
65
64
67 514
Top Syndicated
Programs
Top Syndicated Program in the US since 1997
Rating
Wheel of Fortune (M-F)
Jeopardy (M-F)
Home Improvement (M-F)
Oprah Winfrey Show
Seinfeld
Si
Simpsons
Xena Warrior Princess
Entertainment Tonight
Hercules, Journeys of
Wheel of Fortune (Wknd)
11.0
9.2
8.5
8.0
7.4
62
6.2
6.1
5.7
5.4
5.3
515
248
Audience Metric #4
Gross Ratings Points, Reach
Frequency
q
y
516
Nielsen Media Research
Gross Rating Points
• Rating point
point= 1 percent of
the potential audience
• Gross Rating Points (GRP)
– sum of ratings over a time
period
517
249
• If an advertiser uses four
different programs with
respective ratings of 15, 22,
19, and 27, the weekly GRP
becomes the sum, or 83 GRP
518
4. The Audience Metric #5 or
CUME
• Reach
eac (or
(o CU
CUME))
http://www.all-businesslogo.com/images/update/29aug
2004/Z100__38930.gif
• measures # of viewers or
listeners per week of a channel
• viewers
ie ers counted
co nted once per week
eek
• Useful for cable channels or
519
radio stations
250
Audience Metric #6
Average Quarter Home
Audience (AQH)
• Average audiences for major
time periods of the day
• Shows how many people are
reached over a week
520
Example for CUME:
Radio Station #1
• Station with a CUME of 20,
000 (hi
(high)
h) andd an audience
di
at
an Average Quarter Hour
audience (AQH) of 150 (low)
521
251
CUME: Radio Station #1
Interpretation: station attracts
large numbers of people in a
week but does not keep them
–few listeners at any given time
• Station promotes itself well,
well but
does not have good
programming to keep listeners
522
Example for CUME/AQH
Radio Station #2
• Station with CUME 10,000
(low) and AQH of 2,500 (high)
523
252
CUME: Station #2
–InSmall but loyal audience
–25%
25% of overall listeners
are listening at any
moment
• creases the
h chance
h
that
h ads
d
will be heard by continually
tuned-in audience
524
Audience Metric #7: Average
Frequency (AF) of Exposure
• Used to calculate how many
times an ad must be played so
the average listener will hear
it for example
it,
example, 3 times
525
253
Audience Metric #4
Gross Ratings Points, Reach
Frequency
q
y
526
Nielsen Media Research
Average Frequency (AF)
• AF
AF=AQH
AQH x Number of
Spots Per Week/CUME
• Number of Spots per Week=
{(AF x CUME)/AQH}
527
254
Example for AF: Radio
Station #1
• Assume (AQH=150
(AQH 150, CUME=20
CUME 20,000
000
• To obtain Average Frequency of 3:
{(3 x 20,000)/150}
={(60,000/150)}=400
{( ,
)}
Result: Needs 400 ad spots per week
to reach average listener 3x
528
Example for AF: Radio
Station #2
• Assume AQH
Q = 2,500,
,
,
CUME = 10,000
• To obtain average frequency
of 3 (AF): (3 x 10,000)
/2,500 = 30,000/2500 - 12.
529
255
Radio Station #2
• Need only 12 ad spots
pper week to reach
average listener 3x.
• Will be much cheaper
because more targeted.
But Station 1 will
reach more people
(higher CUME)
http://ww1.prweb.com/prfiles/2005/02/25/2127
79/GManAngleMicTypeshade.jpg
530
Audience Metric #8
Cost Per Thousand
((CPM))
531
256
Cost Per Thousand (CPM)
• the expenditure to reach
1,000 households or persons
with an ad
532
• CPM={(cost
CPM {( t off
advertising)x1,000}/Average
Audience
533
257
Bilotti, Richard, Megan Lynch, Ksenofontova Svetlana “Advertising Outlook
for 2005 and Beyond” Morgan Stanley, 2005
534
CPM for Major Networks
ABC
CBS
NBC
FOX
2000/2001
$18 82 $16.64
$18.82
$16 64 $23.32
$23 32 $16.84
$16 84
2001/2002
$16.59 $17.04 $22.33 $16.96
2002/2003
$17.42 $18.57 $24.12 $17.81
2003/2004
$20.40 $24.31 $29.94 $21.91
Bilotti, Richard, Megan Lynch, Ksenofontova Svetlana “Advertising
Outlook for 2005 and Beyond” Morgan Stanley, 2005
535
258
http://www.morganstanley.co
m/institutional/techresearch/p
dfs/emarketing.pdf
536
CPMs for Various Media
• Prime Time TV
• Radio Network
$16
$6
Magazines (niche) $70 – 190
Magazines (general) $5 – 190
537
259
CPM For Magazines
• Sports Weekly:
$8.75-28.38
• ESPN Magazine: $19.59
$19.59-54.95
54.95
• Sports Illustrated: $19.59-75.17
• Sporting News: $18.71-73.62
• TIME Business Edition: $24.47
• Business Week, Fortune, Forbes:
538
$41.21
http://www.timeplanner.com/planner/editorial/t
argeted_editorial_editions/time
_business_reports_body.html
http://www.usatoday.com/me
dia_kit/sports_weekly/au_eff
icient_reach_men.htm
Cost Per Thousand Impressions
W b Banner
Web
B
Li t Price
List
Pi
CPM (Cost per 1,000
Impressions)
$29
Web Banner Avg. Price
$4
Day Time TV
$5
Direct E
E-Mail
Mail
$20
Solo Direct Mail
$934
Shared Direct Mail
$40
539
260
Different Online Ads
540
• Most newspapers calculate
their CPM as the single column
inch rate divided by their
circulation.
• Magazines
g
determine their
CPM by dividing the cost of a
full page ad by their circulation
541
261
Why Are CPM Prices
Different For Different
Media?
542
1. Different Market Powers of
a Medium
• Different competition in different media
• Local newspapers usually have local
market powers for many types of local
ads.
L l radio
Local
di iis competitive
titi
• New York Times theater box ads: CPM
enormously high
543
262
2. Different Effectiveness of
Media
• Raises willingness to pay
• Based upon length and
quality of exposure,
exposure sensory
involvement, interactivity,
and ease of response.
544
“Cable Advertising Revenue and Addressable Commercials” by Bill Harvey
• The 3-D cube of advertising
value is a way to show average
CPMs for different media
based on three dimensions:
–Targetability
–Sensory
S
iintensity
t it
–Interactivity
545
“Cable Advertising Revenue and Addressable Commercials” by Bill Harvey
263
The Cube of Advertising Value
546
“Cable Advertising Revenue and Addressable Commercials” by Bill Harvey
3. Different Incremental Cost
of Media
• Print media must add paper,
printing,
i i transportation.
i
• TV broadcasting has no
incremental cost per viewer
547
264
Trends In CPM
• For Big 4 TV networks the CPM
increasing, because their value in
reaching national audiences
• For cable:
–Decline for broadbased N/Ws
–Increase for specialty N/Ws
–significant declines for 3rd tier
548
cable networks
Interpretation
• Advertisers lookingg for niche
demographic markets.
• Or, for national reach.
549
265
Station “Rate Card”
• Prices of advertising time
offered
ff d by
b a station.
i
• Includes package plans,
discounts, and policies
• Often starting points for
negotiations
550
551
266
Primetime Ad Prices
(30 sec, US)
Top
Average
1960
$30 000
$30,000
1970
$65,000
1990 $400,000 (Cosby) $125,000
1998 $500,000
$500 000 (Seinfeld)
2003
$455,000 (Friends) $115,799
(http://search.corbis.com/default.asp?i=11328728&vID=1&rID=101)
And Bradley Johnson, Advertising Age, “Low CPM Can Spell Bargain
for Buyers” May 2003
552
267
Media Metric
#9: Quads
554
Nielsen-Type Ratings Measure
Only The Number of Viewer
• It’s a quantity, not the quality
of viewing
• Does not determine the
intensity of preference of
audience.
555
268
• To measure qualitatively, not
just quantitatively, requires
“attitude measurement”
techniques:
–focus groups
–in-depth
in depth interviews
556
Audience Measurement:
“Quads”
• Tool used by TV networks to
study viewing behavior
• 2 factors taken into account:
–tuningg length/episode
g
p
(program’s “holding power”)
–frequency of viewing
(“loyalty”) to program
557
269
Quads distinguish 4 viewer
types
• “Gold cards”:
–watch over 75% of an episode
–Watch over 55% of episodes
shown in analysis period
• “Occasionally committed:”
–watch 75% of program, < 50%
558
of episodes
• “Silver Sliders”
–watch less than 75% of
program, but regularly
• “Viewers Lite”
–watch
watch < 50% program,
program and
rarely
559
270
Advantages
• Holding power indicates
program liking, involvement,
and advertising
–Likely not to switch channels
duringg commercial breaks
560
561
271
• Cable networks have a more
fickle audience than TV
Networks
562
563
272
Audience
Metric
#10 “Q”
#10:
564
“Q”
• Performer is rated on both
f ili it andd how
familiarity
h well
ll s/he
/h is
i
liked
http://www.davidandmaddie.com/images/100tv-people.jpg
Jennie L Phipps. ” Favorites are good buys” Television Week, Apr 21, 2003.
565
273
Performer “Q Score”
• Measure of how much an
audience “likes” a show or
performer
• Evaluations/TVQ Inc., developed
methodology in 1964
Q metric is a derivative of
• “Q”
ratings and overall
recognizability of the star, to
quantitatively assess actors
566
Jennie L Phipps. ” Favorites are good buys” Television Week, Apr 21, 2003.
“Q”
• Q is a ratio of the "Favorite"
score to the "Familiar“ score
• “Familiarity” measures the
proportion of respondents who
recognized the performer
• Respondents also indicate
which stars are their “favorites”567
Jennie L Phipps. ” Favorites are good buys” Television Week, Apr 21, 2003.
274
• This means the Q rating can
be high if a performer is
extremely well-liked
well liked by a
core group
568
Brian Lowry, “Q Marks Spot in the Hunt for What Sells”:. Los Angeles Times. Sep 12, 2001. pg. F.1
James Gandolfini “The
Sopranos”
www.facade.com/celebrity/ James_Gandolfini
• Has a Q score of 36, above
the prime time male average
of 19
Jennie L Phipps. ” Favorites are good buys” Television Week, Apr 21, 2003.
569
275
Q and advertisers
• High performer Q and high
program Q are related
• personality appeal raises a show’s
overall appeal.
• A hi
high
h Q score ffor a show
h often
f
means that viewers watch more
of the commercials
570
Jennie L Phipps. ” Favorites are good buys” Television Week, Apr 21, 2003.
“GYC” Personalities
• GYC programs must have at
least a few identifiable stars
whom the 65+ population like
to watch.
–Mickey
Micke Roone
Rooney
–Oprah Winfrey
571
–Bill Cosby
276
“GYC” Personalities
http://www.africanamericans.com/images2/BillCosbyTimeMag.jpg
572
573
277
574
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
575Need?
• Is This What Media Firms
278
IV. D
IV
Demand
d
Experiments
pe
e ts
576
Demand Experiments
1.
2.
3.
4.
Test Marketingg
Uncontrolled Studies
Controlled Studies
Laboratory
b
Experiments
i
577
279
IV.1. Test Marketing
• Launch the media pproduct
with
i h a full
f ll marketing
k i andd
advertising plan in several
test cities
–Film
–TV show
• Track consumer response
578
Test Marketing
• Problems: Premature exposure
p
of the pproduct to competitors.
• Done for films, with initial
limited roll-out
–Incl. exit interviews
579
280
• Enable decisions about further
development adaptations/finedevelopment,
tuning, and discontinuation.
Aris, Annet, “Value-Creating Management of Media Companies: Chapter 5”,
McKinsey & Company, Inc., 2003
580
Example: TV Show in Small
Country
• The Dutch media producer Endemol
uses the entire Dutch market to test
shows for an international rollout.
Aris, Annet, “Value-Creating Management of Media Companies: Chapter 5”,
McKinsey & Company, Inc., 2003
581
281
IV.2.
IV
2
Uncontrolled
Studies
582
• uncontrolled:
–researchers
are only
observers
b
http://www.unesco.kz/culture/projects/whc/photos/Observers,%20Ms
.%20Kirillova,%20Khorosh%20and%20M.%20Rogozhinski.JPG
583
282
• In contrast, in controlled
research:
–researchers
researchers manipulate the
important variables to observe
their effect.
» more accurate but more costlyy
and time-consuming.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
584
Uncontrolled Research Using Past
Sales Data
1. Aggregate sales data of a
single company
2. Sales data for an individual
retail outlet.
3. Panel data- individual ppurchase
reports from members of a
selected consumer panel.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
585
283
Panel Data
• Marketing research companies
collect individual purchase data
f
from
panels
l off severall thousand
h
d
households.
• Each household keeps a daily
diary of items purchased and
their prices.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
586
Panel Data Advantages
• accumulate observations
more quickly
• One can correlate price
sensitivity
y with demographic
g p
classifications
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
587
284
Panel Data
• Purchases by panel
members can now be
recorded automatically by
in-store POS scanners
- customers could reveal
their
h i demographics
d
hi in
i
return for some store credits
or coupons.
http://www.lib.sfu.ca/about/services/checkout.jpg
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
588
Examples
• Book stores
• Music stores
589
285
Case Discussion: Viacom
“Golden Years”
• Question for Viacom
research: who advertises in
magazines
i
that
th t target
t
t the
th
age group 65 plus?
590
Who advertises to 65+?
• Insurance Companies
– Life
Lif
– Automobile
– Health
– Homeowners’
• Financial services
– Telecom, cable TV, internet
591
286
Who advertises 65+?
• Travel
– Travel agency
– Airlines
– Tour operators
• Pharmaceutical
Ph
ti l drug
d
companies
i
• Food companies
Source: http://assets.aarp.org/www.aarp.org_/articles/benefits/fullbenefits.pdf
592
Golden Years Cable
Marketing
• Golden Years Media may conduct
demand experiments to identify
which products their viewers buy
287
*
Golden Years Research
1. Golden Years Media can obtain
d t about
data
b t their
th i target
t
t households.
h
h ld
Such data can be used to analyze
price sensitivity, etc., with respect
to demographic
g p variables.
595
288
IV.3.
IV
3 Controlled
Studies of Actual
Purchases
596
Experimentally Controlled
Studies of Actual Purchases
• Generate
G
t price
i variations
i ti
while holding constant other
variables, such as
advertising.
d ti i
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
597
289
Controlled Experiments
• buyers are unaware they are
participating in an experiment
• Prices can be varied
• Can also be done for mail-order,
b special
by
i l offers
ff to a subset
b
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
http://www.answers.com/main/content/wp/en/thumb/b/b0/350pxSupermarket_check_out.JPG
598
599
290
In-Store Purchase Experiments
• Such a study can easily
cost several million dollars
• Cost of experimentation is high
because each additional factor
studied requires the inclusion
of more stores as control.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
600
In-Store Purchase Experiments
• For example, when Quaker
Oats conducted an in-store
experiment that focuses on the
effect of price alone, the study
required 120 stores and ran for
three months.
www quaker fr/
www.quaker.fr/
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
601
291
In-Store Purchase Experiments
• Also, charging lower prices
can become too costly for
large-expenditure such as a
TV set or computer
• This leads to the use of
laboratory experiments
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
602
Amazon’s Controlled
Experiment
• Amazon wants to find out
whether a new design of a
webpage increases sales.
• Run a controlled experiment
with a Web page.
Varian, Hal R. “Kaizen, That Continuous Improvement Strategy, Finds Its Ideal
Environment.” The New York Times, February 8, 2007.
292
Amazon’s Controlled
Experiment
• Amazon shows a different page
p g
design to every hundredth
visitor.
• Determination of whether the
new design increases sales can
be made in only a few days.
Varian, Hal R. “Kaizen, That Continuous Improvement Strategy, Finds Its Ideal
Environment.” The New York Times, February 8, 2007.
605
293
IV.4. Laboratoryy
Purchase
Experiments
606
• Using a
research
facility at a
shopping
h
i mall
ll
- simulated
stores the size
of small
convenience
stores.
http://www.we-make-money-notart.com/xxx/FF_150_shoppers2_f%5B1%5D.jpg
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
607
294
Laboratory Purchase
Experiments
• Attempt to duplicate the
realism of in-store
experimentation without the
high cost and exposure to
competitors.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
608
Laboratory Purchase
Experiments
• The researcher controls who
participates and can
manipulate prices etc.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
609
295
• Reward for participating is a
substantial discount
• The cost of laboratory
experiment
i
t is
i muchh smaller
ll
than for in-store testing.
• Popular approach by
consumer electronics makers
http://www.shoplet.com/office/limages/EB021980.gif
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A
Guide to Profitable Decision Making,” Second Edition 1995
610
296
Example for Experiment:
Magazine
g
Test Marketing
g
http://campaignsolutions.com/hdcs/mail/accent.jpg
612
Magazines: Direct Mail
• “Dry Test”
- the product is tested without
being published
- solicitation letters sent out to
potential readers
p
- the first issue may be years away
James Kobak “Testing a New Magazine Through Direct Mail,” How to Start a Magazine
613
297
• Also allows the magazine
company to determine which
combination of design, prices,
offers,
ff
advertising
d ti i copy, andd
mailing lists work the best.
http://www.ptarmigan.co.uk/New%20Pages/DM.html
James Kobak “Testing a New Magazine Through Direct Mail,” How to Start a Magazine
614
• Combining test results with
demographic characteristics
helps a magazine to
determine best target zip code
set, and which other
characteristics to focus on
(Income? Race? Gender?
Optimal Age?)
615
James Kobak “Testing a New Magazine Through Direct Mail,” How to Start a Magazine
298
617
299
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
• Is This What Media Firms Need?
618
V. Measuring
Actual Sales
619
300
Methods of Measuring Actual
Sales
• Books: Bestseller List
• Music: SoundScan
• Film box office
• RFIDs??
620
V.1. Books
Bestseller List
621
301
Bestseller List
• Measured by New
York Times
Times,
Publishers Weekly,
Book Industry
Trends,, Wall Street
Journal, USA
Today
http://images.amazon.com/images/P/044022165X.01.LZZZZZZZ.jpg
http://images.amazon.com/images/P/0451169514.01.LZZZZZZZ.gif
622
623
302
The List is Self-Fulfilling
• Determines book location inside
the store
–Substantial
S bstantial effect on book sales
• Determines whether or not the
book will be discounted
• Compiled from hundreds of book
stores
–identity and weight given to each
store is not disclosed
624
• System
y
is basically
y a very
y big
g
sampling of retailers.
625
303
Manipulating Best-Seller Lists
to Create Audience
• Sampling
p g system
y
of New York
Times Best-Seller list is suspect
• “Padding” the List
–Publishers buyy their own books
in bulk from stores around the
US to get their sales up for the
626
NY Times list
• Business consultants Michael Tracy
and and Fred Wiersema, authors of
The Discipline of Market Leaders,
spent $250,000 to buy 10,000 copies
of their own book, making it a BestSeller. The book spent 15 weeks on
the list
list.
627
http://battellemedia.com/archives/old%20book%206.gif
304
• eventually sold over 250,000
copies.
•NY
NY Times
Ti
now places
l
a
dagger next to any title
when substantial bulk
sales
l are being
b i reported
t d att
individual stores
http://www.majoritynews.com/images/ny-times-logo-paper.jpg
Michael Tracy
628
Fred Wiersema
•
http://ecx.images-amazon.com/images/I/71Q44K6FSCL._SL500_.gif
305
Other’s Best-Seller Book List
• Wall Street Journal offers
“transparency”
transparency of tabulating
sources
–No “weighting”
–reflects
reflects raw sales with no
weight given to any source
• USA Today: point of sale
630
USA hardcover fiction bestsellers 2004
Rank
#
Author
Publisher
# of
copies
Share
The Da Vinci
1 Code
Brown,
Dan
RANDOM HOUSE
3,218,535
19.5%
The Five
People You
Meet in
2 Heaven
Albom,
Mitch
LITTLE, BROWN &
CO PUB
2,065,165
12.5%
Angels &
3 Demons
Brown,
Dan
SIMON &
SCHUSTER
774,668
4.7%
Grisham,
4 The Last Juror John
RANDOM HOUSE
768 609
768,609
4 7%
4.7%
The Rule of
5 Four
Caldwell,
Ian
RANDOM HOUSE
624,956
3.8%
6 State of Fear
Crichton,
Michael
HARPER COLLINS
PUBLISHERS
429,351
2.6%
Title
306
USA trade paperback fiction bestsellers 2004
Rank
#
Author
Publishing
Conglomerate
# of
copies
The Secret Life
1 of Bees
Kidd, Sue
Monk
PENGUIN/PUTNAM
TRADE
865,600
7.0%
The Curious
Incident of the
Dog in the Night2 Time
Haddon, Mark
RANDOM HOUSE
574,294
4.6%
3 The Wedding
Sparks,
Nicholas
WARNER BOOKS
538,139
4.3%
The Lovely
4 Bones
Sebold, Alice
LITTLE, BROWN &
CO. PUB
523,596
4.2%
5 Life of Pi
Martel, Yann
HARCOURT,
BRACE &
COMPANY
522,309
4.2%
One Hundred
6 Years of Solitude
Marquez,
Gabriel Garcia
HARPERCOLLINS
PUBLISHERS
508,381
4.1%
7 The Kite Runner
Hosseini,
Khaled
PENGUINPUTNAM
TRADE
500,338
4.0%
Title
Share
USA mass market paperback fiction
bestsellers 2004
Rank
#
Title
Author
Publishing
Conglomerate
# of
copies
Angels &
1 D
Demons
Brown,
D
Dan
SIMON &
SCHUSTER
2 194 249 13.4%
2,194,249
13 4%
2 Deception Point
Brown,
Dan
SIMON &
SCHUSTER
1,024,273 6.3%
ST. MARTINS MM/
HOLTZBRINCK
1,005,214 6.1%
Digital Fortress: A Brown,
Dan
3 Thriller
Share
4 The Notebook
Sparks,
Nicholas
WARNER BOOKS
671,147
,
4.1%
5 The King of Torts
Grisham,
John
RANDOM HOUSE
654,215 4.0%
6 Bleachers
Grisham,
John
RANDOM HOUSE
516,091 3.2%
Key of Valor: The
7 Key Trilogy
Roberts,
Nora
PENGUIN/PUTNAM
TRADE
489,838 3.0%
307
8 The Guardian
Sparks,
Michael
WARNER BOOKS
9 Blue Dahlia
Roberts, Nora
PENGUIN/PUTNAM
TRADE
431,930 2.6%
10 The Last Juror
Grisham, John
RANDOM HOUSE
399,925 2.4%
Th L
Lake
k H
House
20 The
Patterson,
J
James
WARNER BOOKS
241 921 1.5%
241,921
1 5%
To Kill A
21 Mockingbird
Lee, Harper
WARNER BOOKS
236,337 1.4%
The Catcher in
26 the Rye
Salinger, J.D.
WARNER BOOKS
215,191 1.3%
30 Full Blast
Evanovich,
Janet
ST. MARTINS
MASS
192,373 1.2%
37 1984
PENGUIN/PUTNAM
Orwell, George TRADE
178,699 1.1%
38 Fahrenheit 451
Bradbury, Ray
RANDOM HOUSE
175,725 1.1%
40 Safe Harbour
Steel, Danielle
RANDOM HOUSE
172,281 1.1%
50 Odd Thomas
Koontz, Dean
R.
RANDOM HOUSE
144,808 0.9%
485,649 3.0%
V.2.
2 Music
i Sales
S l
635
308
Music Sales – POS System
http://www.savagebeast.com/images/best-buy-inlines.jpg
636
• Old systems: Selected retailers
(sample) were contacted filled
outt forms,
f
andd returned
t
d them
th to
t
Billboard, Magazine
–reporting often was
inaccurate, merely
rank-ordered
637
–Possible to manipulate
309
Improvement through “POS” [Pointof-Sale] SoundScan System
• SoundScan ((byy Sound Data)) in 1987.
Computerized data collection system
with bar-code scanning by retailers
• SoundScan claims to measure 85% of
all music sales in US.
US
638
http://www.whiteeaglerecords.ca/soundscan-logo.gif
• Point-of-sale purchases are
tabulated from 4,000 chain
record stores, 700 independent
retailers and 7,000 discount and
department stores, and online
stores (`~14,
( 14, 000 outlets in 2003)
639
310
• Billboard magazine uses Sound
Scan since 1991
• Billboard Top Album Lists tracks
the number of units sold and
popularity of particular songs
• Used also by performing rights
organizations (ASCAP, BMI) to
track royalties
640
641
http://www.mixrevolutionblog.com/wp-content/uploads/2007/11/billboard_vinyl.jpg
311
ASCAP Page 676
http://gothamist.com/attachments/arts_jen/2007_08_arts_ascap.jpg
642
• SoundScan owned by Nielsen
• also offers BookScan and
VideoScan
643
312
The Mystery of DVD Sales
• DVD sales information is
important to actors, directors,
and writers for royalties and
profit information.
- distributors usually hype a film’s
film s
initial DVD sales, but do not release
periodic sales information thereafter
John Horn, “DVD sales figures turn every film into a mystery,” Los Angeles Times, April 644
17, 2005, Calendarlive. 15 June 2005.
• In consequence talent agencies
and management firms have
created research teams to
check on DVD revenue and
costs.
p
companies
p
• Or specialized
–Adams Media Research
(AMR)
John Horn, “DVD sales figures turn every film into a mystery,” Los
Angeles Times, April 17, 2005, Calendarlive. 15 June 2005.
645
313
646
V.3.
V
3 Direct Sales:
Measuring Film
A di
Audiences
647
314
Film Ticket Data
• Exhibitor Relations Co.
–Collects
Collects box office attendance
from Studios
–Reports to the media every
week
www.cinedom.de
LA Times, “Tinseltown Spins Yarns, Media Take Bait”, David Shaw, Feb 12, 2001
648
Film Audiences
Sunday (am) –
theatres report
Fr/Sa ticket sales
Media - Monday Box
Office Report
Company collects
info from studios,
and reports to
media
Chosen theatres
in key markets
Studios
extrapolate
Fr/Sa data to
guess Su
Exhibitor Relations Co.
Co
Extrapolate for smaller
markets estimate
649
LA Times, “Tinseltown Spins Yarns, Media Take Bait”, David Shaw, Feb 12, 2001
315
Film Box Office Weekly Report
Weekend Top 30 Box Office
650
Movie Reporting Criticism
• Potentially Inaccurate
–The numbers are “made
up”—fabricated every
week” (Anne Thompson,
week
Thompson
editor, Premiere magazine)
LA Times, “Tinseltown Spins Yarns, Media Take Bait”, David Shaw, Feb 12, 2001
651
316
Movie Reporting Criticism
• Potentially manipulative
–The studios extrapolate
p
the
Sunday figures from the
Friday-Saturday figures, based
on experience.
–Want to have the number one
movie of the week.
–Exaggerate, to drive future
sales
652
LA Times, “Tinseltown Spins Yarns, Media Take Bait”, David Shaw, Feb 12, 2001
- To make sure theaters are
not misreporting the number
off ti
tickets
k t sold,
ld distributors
di t ib t
employ undercover checkers,
who buy numbered tickets at
th first
the
fi t andd last
l t shows
h
att
randomly selected theaters.
Epstein, Edward Jay, “The Big Picture, The New Logic of Money and Power in
653
Hollywood,” New York: E.J.E. Publications, Ltd., Inc., 2005
317
Direct Sales Data
• Film studios also receive direct
information from national and
regional multiplex chains in the
United States and Canada.
http://www.gjdc.org/images/Multiplex%20Cinema.jpg
Epstein, Edward Jay, “The Big Picture, The New Logic of Money and Power in
654
Hollywood,” New York: E.J.E. Publications, Ltd., Inc., 2005
• Studios also conduct exit
polls, to determine the
demographics of audiences.
Epstein, Edward Jay, “The Big Picture, The New Logic of Money and Power in
655
Hollywood,” New York: E.J.E. Publications, Ltd., Inc., 2005
318
• Nielson National Research
G
Group
((NRG)
G) main
i tooll for
f
film audience research but
others were catching up.
Dutka, Elaine. “Audience Tests: Plot Thickens.” 31 August 2003. Los Angeles Times. Last656
accessed on 4 June 2007.
V.4. RFID
Tracking
g
657
319
More Refined Tracking: RFIDs
(Radio Frequency Identification
• As passive
p
(unpowered) RFIDs tap
prices come down to
ppennies,, it is on the
verge of becoming
major measurement
658
tool
http://www.pdcorp.com/healthcare/photos/chip_hand.jpg
http://www.elektroniknet.de/topics/kommunikation/fachthemen/2003/0021/images/3190908_kl.jpg
RFID:
• A ppassive radio transponder
p
with view-ware that reflects
an integrating radio signal
received
659
320
RFID
• The RFID tag is a small integratedcircuit chip with a radio and
identification code embedded into it,
which can be scanned from a
distance.
• likely to replace barcodes.
660
RFID in tracking merchandize
•In 2005, Wal-Mart required its top 100
suppliers to apply RFID labels to all
shipments so as to improve supply chain
shipments,
management
•Next step to tracking at POS with potential
ID and profiling of use potential to
consumer’s home.
•Research tool for real time audience analysis
Source: IEEE Computer Society, RFID: A Technical Overview and Its Application to the Enterprise
http://doi.ieeecomputersociety.org/10.1109/MITP.2005.69
661
321
RFID
• Samsung developed RFID fridge:
- suggest recipies based on what
you have in fridge or compiling a
shopping list…
• Same idea could be used for
music CDs- suggested play list
for the evening
–Could be linked to media company
for audience analysis
662
Tracking “Best of Golden
Years” DVDs
• An RFID tagg will enable
“Golden Years” to track every
individual DVD purchased.
• This allows an accurate
measurement of all sales.
663
322
664
VII. SelfReporting
665
323
VII.1. Measuring
Circulation
• Producer Self-reporting
• Circulation Verfication
• Problems
P bl
with
ith Measuring
M
i
circulation
666
A. Producer SelfReporting
• Mainly used by newspapers
newspapers,
magazines
• Each media company sends
reports on circulation, ad sales
andd other
th relevant
l
t information
i f
ti
to a central unit
667
324
Producer Self-Reporting
• The central unit compiles the
information and prepare
different reports
• The central unit also
responsible for auditing
668
Central Self-Report
Model
Magazine Z
Advertisers
Reports
Central
Unit
Magazine W
Reports
Magazine Y
Specified Data
(Circulation, ads, etc.)
Magazine X
669
325
Audit Bureau of
Circulation (ABC)
• Began in 1914
• formed to audit and verify
circulation
• Before ABC, advertisers had to
face boasts about sales
sales.
• Led to overprinting and dumping
• Advertisers and ad agencies
create ABC to sort the mess 670
ABC Board
• 12 advertiser and ad-agency
directors
• 6 daily newspaper directors
• 3 magazine directors
• 1 director representing weeklies,
farm publications, business
publications and Canadian
671
periodicals
326
ABC process
• Half yearly, newspaper
members supply publisher’s
statements that detail how
and where each copy sold.
• Once a year, ABC audits
sales
672
Publisher's Statements
• Twice a yyear,, ABC requires
q
each magazine and newspaper
member to submit a statement
of their circulation -- known as
a Publisher's Statement.
http://www.accessabc.com/aboutabc/index.htm
673
327
Sample ABC Report
674
•Newspapers also conduct
telephone surveys(sampling)
–Simmons, (large consumer
research firm), conducts
newspaper reader research
675
328
Problems with Measuring
Newspaper Readership
•
•
Information about section or even
story readership difficult to obtain
Demographic
information not
part of selfreporting
676
Issue: How to Define
Circulation?
677
329
Newspapers “circulation”
• Circulation = ppaid subscriptions
p
+ newsstand sales
http://www.michaeljacksontalkradio.com/Journals/MJs_Journal04_0317.htm
678
• How to count bulk copies to
hotels, businesses, hospitals?
–How steep can discounts be?
679
330
• The ABC specifies that a paper
must be sold for at least 50%
of its normal pprice to be
counted as paid circulation.
680
http://www.experientia.com/blog/uploads/2007/03/usa_today.bmp
681
331
Newspapers and Third-Party Sales
• Problems with counting papers
distributed for free by 3rd parties
• Over third-party sales to buys
by external companies that
distribute them for free ((e.g.
g
hotels, airlines)
http://mowabb.com/aimages/archives/003933.html
JACQUES STEINBERG AND TOM TOROK, “Your Daily Paper, Courtesy of a Sponsor,” The New York Times,
682
January 10, 2005, C6
• Excluding third-sales the
average paid circulation of
USA Today and The Wall
Street Journal would have
dropped 2%.
JACQUES STEINBERG AND TOM TOROK, “Your Daily Paper,
683
Courtesy of a Sponsor,” The New York Times, January 10, 2005, C6
332
Mis-Reporting of Circulation
Numbers
• 2004: Belo Corp.
p ((Dallas Morning
g News
and other papers, and 19 TV stations)
–Investigation on false numbers
–Counted unsold papers
–Overstated circulation 5.1%, Sundays
11 9%
11.9%
• Refunds $23 Mil, loses advertiser
confidence
684
Belo Corp.
http://cache.daylife.com/imageserve/07kf7XU5UEcuB/610x.jpg
685
333
http://www.billnealonline.com/siteassist_images/DMNews.jpg
686
Newspaper Circulation
http://www.cartoonstock.com/directory/c/circulation.asp
687
334
Mis-Reporting of Circulation
Numbers
• Other mis-reporting newspapers:
–Hollinger (Chicago Sun-Times)
–Tribune Co. (Newsday, Hoy, etc.)
–Counted unsold copies not
returned
–Criminal investigation
–Overstated 40,000 copies,
688
Sunday, 60,000 copies
http://sadbastards.files.wordpress.com/2006/11/sun-times-small.jpg
689
335
http://www.dyingwell.com/images/newsday.jpg
690
Redefining a Paid Paper
Many of the country’s largest newspapers have been counting
papers paid for by a third party, like an advertiser, as part of
their paid circulation. Here are some of the larger newspapers,
ranked by use of third party sales
Newspapers with circulation
of 250,000 or more
Publisher
Total paid
Circulation, six
months, ended March
2004
Third Party Sales as
percentage of total paid
circulation, 2004
USA Today (Fridays)
Gannett
2,635,412
18%
The Denver Post
MediaNews
783,274
13.2%
The Wall Street Journal
Dow Jones & Company
2,101,017
8.4%
The San Jose Mercury News
Knight Ridder
308,425
8.3%
The Houston Chronicle
Hearst Newspapers
p p
740,005
,
8.2%
The Miami Herald
Knight Ridder
447,326
6.8%
The Philadelphia Inquirer
Knight Ridder
769,257
5.8%
The Boston Globe
The New York Times
686,575
4.4%
The Harvard Courant
Tribune
283,410
4.0%
Los Angeles Times
Tribune
1,392,672
3.8%
691
The New York Times, 10 January 2005.
336
Alternatives to ABC
•BPA
BPA International
(business magazine in
20 countries)
•Mediamark Research
(consumer magazines)
692
• Other magazine circulation
reports:
–Folio
Folio 400 tracks newsstand and
subscription sales of top 400
magazines
–Magazine
g
Publishers off
America - track circulation for
its 200 member magazines and
693
periodicals
337
New Problems: Multi-Platform
• How to measure audiences that use
multiple platforms?
–paper newspaper & online
li paper
–radio station over-the-air and online
Some online are the same people not
additional ones (for most newspapers
about 15% of visitors are not paper
subscribers).
ABC (Audience Bureau of Circulation)
2006 new “consolidated” product 694
695
338
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special
Problems of Media Demand
Estimation
• Case Discussion: Viacom
Golden Years Media
II. ANALYTICAL/STATISTICAL
MODELS
• Statistical Inference
• Econometric Demand
Estimation
• Conjoint Analysis
• Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of
Actual Purchases
• Laboratory Purchase
Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
VII. SELF-REPORTING
• Sampling Methods
• Next Generation People
Meter: The Digital Meter
I.
• Auditing
696
VIII CONCLUSIONS
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
697Need?
• Is This What Media Firms
339
VI.
Measuring
Traffic
698
VI.1.
3 Approaches to
Measuringg Internet
Audiences
699
340
Top Websites to US Internet Users
for April 2008
Rank
1
1.
2.
3.
4.
5.
6
6.
7.
8.
9.
10.
Website
Google Sites
Yahoo! Sites
Microsoft Sites
AOL LLC
Fox Interactive Media
eBay
B
Wikipedia Sites
Amazon Sites
Ask Network
Time Warner –Excluding AOL
Unique Visitors (000)
141 080
141,080
140,613
121,213
111,277
87,527
80 903
80,903
58,812
58,057
54,086
700
52,544
How do we
know that?
701
341
Approaches to Measuring
Internet Audience
A. Site-Level
B. Ad-Level
C User-Level
C.
User Level
702
3 Approaches to Measuring
Internet Audiences
A. Site-Level
– Count website visits. Similar to actual
sales approach
B. Ad-Level
– Measuring clicks on ads when user is
transferred to advertisers
advertisers. Similar to
actual sales approach
C. User-Level
– Built by 3rd parties from panel/meter
data, similar to TV ratings approach 703
342
A. Site-Level
Measurement
http://kentaro.blog.ocn.ne.jp/kentarob
log/images/yahoo-search-thumb.jpg
http://www.politicalpuzzle.org/Photos/msn%20se
arch.jpg
http://news.bbc.co.uk/1/hi/business/1476
805.stm
704
Site-Level Measurement
• Basically, a self reporting
system
t by
b th
the website
b it or visitor,
i it
• Can potential identify users /
user types/countries, etc.
• Tabulations
b l i
off page requests
• Most commonly used by
websites
705
343
Uses for Internet Ratings
• Total website hits can be used
as the
th basis
b i for
f determining
d t
i i
unique users, given a
relationship between the two.
• Best fit: modified exponential
function:
706
[
UniqueUsers = 3.2 1 − e (.004599−.090583*Hits )
]
707
344
Internet Measurement
Software
708
http://www.vioclicks.com/pics/signupbig.gif
Nielsen’s Ne Ratings software:
SiteCensus:
• Nielsen//Net Ratings (2003)
• Browser-based
Bro ser based measurement
meas rement tool
• Makes variety of data available to
media owners
• Paths followed
• Content
C
viewed
i
d
• Location of access
• Includes requests from work,
school, and wireless
709
345
Server Level Collection
• “Packet sniffing”
–Monitors network traffic
coming to a website and
directly extracts usage data
from TCP/IP packets.
Jaideep Srivastava, Robert Cooley, Mukund Deshpande, Pang-Ning Tan,
“Web Usage Mining: Discovery and Applications of Usage Patterns from
Web Data” in SIGKDD Explorations, Vol. 18: 2, January 2002 p. 13.
710
http://www.krittersinthemailbox.com/
animals/dogs/bloodhound/sc1139.htm
Site-Level’s Systematic
Measurement Biases
• Overcounting
– repeat visitors
– counts not just people but also bots
and spiders
• Undercounts cached pages
• Can’t
C ’t distinguish
di ti i h multiple
lti l users on off
same computer
711
346
Problems with Site-Level
• Knows IP address or technical
details not user identity.
details,
identity
http://www.montanahope.org/graphics/bears%20and%20computers.JPG
712
How to Individualize
Information about a Web-site’s
Audience
• Registration requirements do
not work well
– Effort to users
– Privacy concern
– fear of spam
713
347
Major Tool: Cookies
• Cookies combine the control advantages of a
site-centric
site
centric approach with the
individualization of the user-centric approach
• A standard programming device that
produces electronic files to tag individual
customers with a unique identification
identification.
– Allows a website to recognize an
individual.
Deck, Cary A., “Tracking Customer Search to Price Discriminate.” Electronic Inquiry,
714
April 1, 2006.
715
348
B. Ad-Level
Measurement
716
http://www.smarteque.com/
Click-Through (CTR)
Software
• Measures whether user clicked
on an ad to link to the advertiser
717
http://www.answers.com/main/content/wp/en/thumb/0/03/325px-Pop-up_ads.jpg
349
Click-Through (CTR)
Software
• Valuable to advertisers:
measures actual effect of web
advertisement; unique to
Internet
• Some
S
per-click
li k payments
t quite
it
high--$20!
• Usually < 1$
718
Inflated Click Rates
• Creating fake clicks
• robot hits
• This has become a big
pproblem
• Fake clicks by people
has become a cottage
industry in India
http://ewic.bcs.org/images/robot.jpg
719
350
• Major Abuses of Pay-PerClick:
–“Click fraud” not illegal
g
–Portals like Yahoo have
disincentive to crack down,
incentive to click fraud, through
sharing of PPC that are charges to
advertisers
d ti
–Attempts for techno-fixes have
failed
720
721
351
C. User-Level
Measurement
722
http://www.infosystem.gr/images/computer_user3.jpeg
User-Level Measurement
• A Sampling technique
• Drawn from TV audience sampling model
– Large panel of randomly selected users
– Software meter on user’s PC measures
behavior
– Meter reads the URL in the browser,
counts and forwards data to web-rating
counts,
company
Source: Scott MacDonald
723
352
Data Processing
• The data are matched to
“dictionaries of the Internet,”
which categorizes the millions
of recorded URL’s
Steve Coffey, “Internet Audience Measurement: A Practitioner's View,” in
724
Journal of Interactive Advertising, Vol. 1:2, Spring 2001, p. 13.
725
http://www.mediasmart.org.uk/images/photos/girl_on_computer.jpg
353
Advantages of User-Level
Approach
• Uniform measurement -->
comparability
• provides demographics
• Counts pages actually received
• Measures actual behavior (not
self-reported)
• No conflict of interest
726
• Requires user
cooperation.
• Incentives are
offered to users
who are willing
to use the
browser.
http://www.heart-disease-bypasssurgery.com/data/images/incentive.gif
Jaideep Srivastava, Robert Cooley, Mukund Deshpande, Pang-Ning Tan, “Web
Usage Mining: Discovery and Applications of Usage Patterns from Web Data” in 727
SIGKDD Explorations, Vol. 18: 2, January 2002 p. 13.
354
The Data Meter
• In 1995, Media Metrix installed
the first meter of internet uses,
the “PC Meter,” into a consumer
sample
http://www.queensferry-pri.edin.sch.uk/nursery/photos/computer2.jpg
http://www.netprointer.com/image_file/seo_image/image021.gif
Steve Coffey, “Internet Audience Measurement: A Practitioner's View,” in
728
Journal of Interactive Advertising, Vol. 1:2, Spring 2001, p. 11.
Web Ratings War
• Nielsen,, a news monopolist
p
in
TV ratings but not in web
ratings
– 100 web ratings companies,
such as comScore, Hitwise
Johnnie L. Roberts, Newsweek, Nov 27, 2006
729
355
Web Rating Companies
(Nielsen)
Source:Web rating: Heavy traffic ahead, The Industry Standard 9/18/00
730
Methodology
• Sample randomly recruited by
phone and mail. Sample of
50,000.
731
356
Problems with UserCentric Measurement
• Disadvantages to small sites
which may get only a few hits
and may be ignored or
undercounted
• Poor site diagnostics (no good
info on sites and what user does
there)
732
Cookies
• Online retailers can use
cookies to post dynamic,
customer-specific prices.
Deck, Cary A., “Tracking Customer Search to Price Discriminate.”
Electronic Inquiry, April 1, 2006.
733
357
VI.2. Data Mining
734
Data Mining
• The Internet also
provides a powerful
tool for additional
analysis
• The capacity to track
users’ browsing
users
behavior
http://www.nada.org/Images/Technology_image3.gif
735
358
Mouse Activity
- number of clicks
- time spent
p movingg
the mouse in
milliseconds
- time spent scrolling
http://www.dalveydepot.com/DalveyBMS.jpg
Mark Claypool, David Brown, Phong Le, and Makoto Waseda, “Inferring
User Interest,” in IEEE Internet Computing, Vol. 5:6, November 2001, p.
736
35.
Study Results
• Total time spent on a Web page
and total time spent scrolling the
mouse is a reliable indicator of
interest.
• The
Th number
b off mouse clicks
li k is
i
not a good indicator of interest
Mark Claypool, David Brown, Phong Le, and Makoto Waseda, “Inferring User
Interest,” in IEEE Internet Computing, Vol. 5:6, November 2001, p. 37.
737
359
Web Usage Mining
• Demand of internet sites can
b measuredd using
be
i webb usage
mining.
• This process is a data mining
technique used to find the
usage data
d t off webb sites
it so
web applications can be used
better.
Srivastava, Jaideep, Cooley, Robert, Deshpande, Mukund, & Tan, Pang-Ning. “Web Usage
738
Mining: Discovery and Applications of Usage Patterns from Web Data.” SIGKDD Explorations.
1,
no. 2 (January 2000):12-22.
Web Usage Mining
• Pattern discovery is the usage of
algorithms
l ith to
t find
fi d usage patterns.
tt
Srivastava, Jaideep, Cooley, Robert, Deshpande, Mukund, & Tan, Pang-Ning. “Web Usage
739
Mining: Discovery and Applications of Usage Patterns from Web Data.” SIGKDD Explorations.
1,
no. 2 (January 2000):12-22.
360
Case Discussion:
How to Measure the
Usage of the “Golden
“
Years” Internet Portal?
741
http://www.thrombosis-charity.org.uk/support.htm
361
How Do We Know
How Many Internet Users
“G
Golden
ld Y
Years” Attracts?
Att t ?
How Many
U
Users
R
Read
d
Its Ads?
742
User-Centric
• Obtainingg data from userlevel method of measurement
would be helpful. But user
ppanels probably
p
y do not cover
GY’s older demographics
well
743
362
Ad-Centric
• Measuring Ad-clicks/hits from
GY’ss website to advertising
GY
sites helps Golden Years Media
in two ways:
–Raises
Raises advertising revenues
–Provides information on what
interests the visitor.
744
Site-Centric
• A website “hit” counter can
collect data on the number of
hits/clicks to GY Portal to
measure demand for the website.
Together with cookies,
cookies this
would provide good information
about GY’s online audience.
745
363
746
I.
OUTLINE: MEDIA DEMAND
ANALYSIS
WHY DEMAND ANALYSIS IV. DEMAND EXPERIMENTS
• Importance and Special Problems
of Media Demand Estimation
• Case Discussion: Viacom Golden
Years Media
II. ANALYTICAL/STATISTICAL
MODELS
•
•
•
•
Statistical Inference
Econometric Demand Estimation
Conjoint Analysis
Diffusion Models
• Test Marketing
• Uncontrolled Research
• Controlled Studies of Actual
Purchases
• Laboratory Purchase Experiments
• Psycho-Physiology Testing
V. MEASURING SALES
•
•
•
•
Books: Bestsellers
Music Sales
Film Audiences
RFIDs
III. EMPIRICAL SAMPLING OF
AUDIENCE/ CONSUMERS VI. MEASURING TRAFFIC
• Sampling Methods
VII. SELF-REPORTING
• Next Generation People Meter:
The Digital Meter System
• Metering Alternative: Cable Box
and Tivo Box
• Audience Metrics
• Qualitative Measures
• Auditing
VIII. CONCLUSIONS
747Need?
• Is This What Media Firms
364
VIII
VIII.
Conclusions
748
Tools
Covered
749
365
In this Chapter, we covered the
following Analytical (not
technical) Tools for Demand
Estimation:
• Statistical inference and
sampling
• Delphi
D l hi andd Comb
C b analysis
l i
• Audience model-building
• Econometric demand estimation750
Tools (cont.)
• Construction of Upwardsloping
p g demand schedule
(Network effects)
• Design of surveys
• Paretian revenue distribution
• Conjoint Analysis
• Epidemic models of diffusion 751
366
Tools (cont.)
• AQH, AF, Qumes audience
metrics
• Relation of ad revenues to
macro-economy
• Controlled Experiments
• Panel data use
• Internet surveys
752
Tools (cont.)
• Psycho-physiological
y
p y
g
techniques
753
367
Issues
Covered
754
Issues
• Nielsen & Arbitron
methodologies
• People meters and PPV
• POS measurement
• Self-reporting methodology
• Click-counting
755
368
Issues
• Statistical estimation of demand
• Forecasting methodologies
• Internet methodologies
• Etc.,
Etc etc.
etc
756
Issues
• Special
p
Problems of Demand
Estimations
• Analytical & Statistical Models
• Econometric Models to Estimate
Demand and Related Problems
• Problems of Diffusion Models
757
369
Issues
• Nielsen & Arbitron methodologies
g
• Measure Internet Traffic: site-level
measurement, user-level
measurement, and user-centric
measurement
• Internet Self-reporting
758
Issues
• Special
p
Problems of Demand
Estimations
• Analytical & Statistical Models
• Econometric Models to Estimate
Demand and Related Problems
• Problems of Diffusion Models
759
370
760
Case Discussion
Viacom “Golden Media”
Should Viacom survey potential
viewers? How?
761
371
Case Discussion:
Econometric Estimation
• “Golden
“G ld Years”” VOD
O
–What price to charge?
–Need to find pprice elasticity
y
of consumers
762
• Need to specify a “model” for
statistical estimation
• Example:
• Q is the total number of VODπ
orders by subscribers
763
372
“GYC: Historical
Analogy
• “Golden
Golden Years”
Years can forecast
GYC’s market penetration by
analyzing the growth of a
similar channel.
channel
764
• And if there is no channel
dedicated to people 65+, it may
be possible to make estimations
based on the growth of a channel
targeted for a specific population,
such “Lifetime”
Lifetime television for
women, the N (teens), Spike TV
(men), Logo (gay), or BET
765
(African-Americans)
373
• We asked the questions
- how can Viacom determine
d
demand
d and
d related
l d
information for still nonexistent products
766
• Demand for its still now existent
products
• Characteristics of viewers/readers
• Willingness to pay
• Characteristics of non-buyers
• Interest by advertisers
• How to portion its products
• How to plan marketing strategy
• How to plan pricing strategy
• What the audience likes/dislikes about
767
374
• We now understand better the
ppotential actions and their
effectiveness.
768
To predict the audience for the
GY cable channel
• Early planning: - personal surveys
- Focus groups
- Conjoint analysis
- Delphi Surveys
- Diffusion studies
769
375
Planning
• Content stage
g
- Focus groups
- Test marketing
- Psycho-physiological
h h i l i l tests
770
Once channel is running
• Phone surveys
y
• People meters (if audience is large
• Cable box
• Econometric studies
771
376
For attracting advertisers to
audience
• phone surveys of viewers
• controlled marketing research
p ads
for impact
772
For Golden Years Magazine
• Same as for GY Channel, thus
achieving synergies
- add: direct mail test grid survey
- add: actual rates data
- add: surveys of actual subscribers
- drop: people meters and cable box
773
377
For the Website “The GY
Postal”
• Use some of the same information
• Add: cookies (on user PC)
• Add: click data (on ads)
• Add: data on visitors (website)
(
)
774
• We can see that there are a large
number of approaches to collect
data
• In near future, the tools of online
and video tracking will permit a
real-time matching of audience,
i l di the
including
th choices
h i
off nonviewers in the target
demographics
775
378
•
•
•
•
Thus, strength in data collection
But how is the data used?
This is the weakness: research follow on
C
Current
t methodologies
th d l i are pretty
tt impractical
i
ti l
- econometrics (need data, must project past
into future)
- identification of references by sociodemographics
- epidemic model projections
- trade-off (conjoint)
776
• No strong link to behavioral
models and analysis (psych,
sociological,
g , behavioral
economics)
• This is the challenge
– not just more data
–But more advanced “data
mining”
777
379
So we covered a lot of ground.
But a last and important
question remains beyond
techniques, and technologies,
and technocratic
management: whether such
techniques are really what
media firms need
778
Should media companies
use demand estimation
techniques, like a car
manufacturer or an
airline?
i li
779
380
1. Should One Avoid
Forecasting on Practical
Grounds?
• Many are inclined not to
forecast at all before
launching a media product
because forecasts are so
Carey, John & Elton, Marin. “Forecasting demand for new consumer services:
inaccurate.
challenges
and alternatives.” New Infotainment Technologies in the Home. Demand780
Side Perspectives, Lawrence Erlbaum Associates, New Jersey, pp. 35-57.
Critics of MBAs in News Media:
• “It is a fantasy to believe that a
newspaper can be designed and
packaged like a bar of soap or a
can of dog food or even like a
television news program.
program ”
–Leo Bogart, retired executive VP of the
Newspaper Advertising Bureau
Doug Underwood When MBAs Rule the Newsroom: How the Marketers and Managers
Are Reshaping Today’s Media. New York: Columbia University Press, 1993, pp. 3-13.
781
381
The Limits of Conventional Research
for Newpaper Audience
• No longer viewed as a Panacea for
circulation problem;
• Often mere restatement of common
sense at the most
• OftenCommunication pproblem between
researchers and decision makers
•Need for theoretical models editors can
follow
Philip Meyer “Limitations in Conventional Newspaper Research” The Newspapaer Survival Book, 782
Bloomington: Indiana University Press
The Limits of Conventional
Newpaper Research
783
382
Entertainment
• Disney ex-CEO Michael Eisner:
Research is good on past or present,
not on future.
• Audience
wants
originality,
up to a
point.
784
http://www.azcentral.com/arizonarepublic/news/gifs/0911eisner.jpg
2. Should Media
Companies Go Beyond
Short-Term Efficiency?
785
383
•Do media owe its audience a
special responsibility to go
beyond what its audience
wants ?
- unpopular news stories
- breaking taboos
786
Should one Avoid Measurements
on Principled Grounds?
Time,, Inc. Former
Editor-in-chief Norman
Pearlstine:
Balance between seeing
readers
d what
h they
h want,
and what we think they
need.
http://image.pathfinder.com/fortune/conferences/globalforum/625.jpg
787
384
“There’s always been a balance
between educating your reader
and serving your reader… you
obviously balance telling them
what you think they ought to
read with giving them what they
want to read…”
788
Recall the earlier question:
• Does the audience’s demand
shape the content supply?
• Or does supply—by large
media firms
firms—shape
shape viewer
preferences and demand?
789
385
• Are media demand-driven?
–As much of the audience
research techniques imply?
• Or are they supply-driven?
supply driven? As
marketing activities imply?
790
• As often the case, both side are
partly right.
• Advertising,
Ad ti i PR,
PR andd media
di
content itself shape public
• But audiences also reward
originality, and many do not
want to be pandered.
791
386
• Creativity required not only in
the media product itself,
•But also in understanding the
audience’s needs, tastes,
preferences, desires, fears.
•These demand factors are often
subconscious, unarticulated by
792
audience
So, is demand analysis
• “bean-counting” by uncreative
minds
• Tool for pandering to audiences
rather than of leading them?
793
387
• A manager should not make the
choice between judgment and
empirical estimation.
• Used effectively, they are
complementary.
Thomas T. Nagle & Reed K. Holden, “The Strategy and Tactics of Pricing: A Guide794
to Profitable Decision Making,” Second Edition 1995
• The avant-garde media manager is
3 steps ahead of audience
•Conventional media managers
follow the audience by one step,
letting audience research make their
decisions
•The moderately successful media
manager: probably one step ahead,
using audience research
795
388
• The successful innovator: 2
steps ahead, creative
understanding of audience,
market, and society, plus
research to lower the risk
796
797
389
To Conclude:
• Determiningg and analyzing
y g
demand for media is
increasing in its technological
sophistication
p
798
• We now have new technical
tools:
–Internet connectivity for media
consumption
ti
–Local People Meters
–Measurement software
–Cookies
C ki
–RFID
–Watermarks and IDs
799
http://www.smwinc.com/news/img/03wn/rfid.jpg
390
• These tools provide
enormously powerful
methods of instant
feedback
800
• Thus, demand measurement of
media use will be increasingly
–real-time
–global
–large
g samples
p
–customized
http://images.google.com/imgres?imgurl=http://210.75.208.159/eolympic/xbj/txtx/image/txtx.jpg
801
391
• But even with these better
tools,, it is much harder to do
demand research today
802
• It is harder to estimate
demand for new products and
services in a rapid-change
rapid change
environment, with
fragmented audiences, and
much greater choice,
choice and
shorter attention spans
803
392
• Media firms will increasingly get
rapid audience data and act
rapidly on them, in the design of
th i products,
their
d t in
i marketing,
k ti andd
in pricing
804
• As sophisticated as the tools are
which have been reviewed, they are
probably just beginning of to develop
the next generation of tools utilizing
much more advanced
–Behavioral research
–Audience instant feedback
–Trendsetters
–Cross
Cross cultural
c lt ral sampling
–Statistical tools
–Online technology
805
393
Demand Analysis Becomes
More Important
• The greater the uncertainty
• The greater the upfront
investment
• The ggreater the economies of
scale and network effects
• The more competitive alternatives
806
• The shorter the product cycle
• Reliance on the “gut feeling”
“intuition” of “single-minded
entrepreneurs and of internal
advocates can be the most
expensive way to learn.
807
394
• Suppose a film has
Cost = $50 mil.
(Probability) P = 20% to gross $250 mil.
(Expected Return) E (R) = .2 x 250 = $50 mil.
E (Profit) = $50 mil cost- $50 mil [E (R)]= 0
• If one can improve the odds from 20% to
22% by smarter demand analysis.
E (Profit) = .22 x 250 = $55 mil
ΔE (Profit) = $5
$ mil
• Now profit expectation is positive
808
And therefore, I
disagree with
ith the
slogan that “Nobody
Knows Anything”
Anything .
809
395
One can improve the odds
• Slightly,
g y, but that is enough
g
for a competitive advantage
810
“Somebodyy Knows a Little
Better”
811
396
• Understanding One’s Audience
may be cheapest investment
with the highest return.
return
812
• And Demand Analysis—
y
understanding the audience,
customers, market, is the key
to improve
p
the odds.
• We are just at the beginning.
813
397
End of Lecture
815
398