Star Gazing – Why Bother?
Transcription
Star Gazing – Why Bother?
“Star Gazing – Why Bother?” Ken Diefenbach Position Title : Project Manager, Business Intelligence @ CQU Information Technology Division Abuse, Arguments, Questions, etc: k.diefenbachATcqu.edu.au 11/11/2007 1 Presentation Sequence Audience Poll Setting the Scene DW 101 What’s Wh t’ progressed d since i 2006 Current Situation Future Directions Questions / Comments 22/09/2007 2 Setting the Scene - BIG CQU – COGNOS site since early 90’s 90 s for Business Intelligence – PowerHouse site since before I started there in 1989 Business Intelligence Group – ((“ Think BIG”) BIG ) (“Think BIG” – Initially 3 members, 90+ years IT experience As of Friday last, 5.7 EFT And we lost 30+ of those years of experience Moi – 15 years + in Business Intelligence – Chair of HEUG Reporting & Analysis Product Advisory Group Group - Prone to talk too much and too fast - Please tell me if it’s happening again! •NB This makes it YOUR FAULT if you let me rabbit on!!! 22/09/2007 3 Data Warehouse 101 NOT a copy py of the transaction data base on a separate server A repository where information from multiple sources is restructured and stored in a format more suitable for analysis and reporting CQU is i still till adopting d ti th the W Warehouse h Gospel a la Ralph Kimball – Operational p Data Stores – Star Schemas Sometimes the we disagree team disagrees on on interpretation or emphasis – but being the boss helps ☺ 22/09/2007 4 DW 101 or – Why Stargaze? Stargaze?… Not because it’s it s fun Not because almost everyone is doing it Hang it on some business needs! Faster reporting from the data warehouse Faster access to the originating applications Better quality data 22/09/2007 5 Why Stargaze Part 2? Value added data – information that does not exist in the source applications Ability to report across applications Provide a single version of the truth Have the ability to import additional data Provide better data structures for ad hoc reporting 22/09/2007 6 Limitations Not necessarily the latest view of the data Development will be evolutionary, not big bang Data must already exist or be derivable – Within ERP – Within Excel / Access CSV / text 22/09/2007 7 Some appropriate music please Maestro…. Maestro http--www.wdisneyw.co.uk-photos-wish.mid.url 22/09/2007 8 Kimball According to Disney Base your report on a star Makes no difference who you are Fact tables and dimensions Will work for you… Demands from Chancellery or Dean? No request is too extreme! Anything their hearts desire Can come from you…. 22/09/2007 9 Dimensions, Dimensions Facts Student Term Admissions reporting Admissions 22/09/2007 Leave type yp Job Action Time Career Application Staff Leave reporting Faculty Leave 10 What do you need to get a ? warehouse? A location to store it – “Pontious” “P ti ” was our pilot il t – Living proof that SysAdmins can’t spell – And / or never went to Sunday School A way to get the data in – An ETL tool is one way: EXTRACT the data from multiple sources TRANSFORM this data into a different form to optimise it for reporting LOAD it into the Data Warehouse – And you need a server to run this on “Morph” is the server that we use to massage and reshape the data A window to run the “refreshes” in according to need – Refresh our complete student data early each morning except Sunday (ODS and Warehouse) – Takes pretty much 5 1/2 hours from 2am – 7:30 am counting PowerPlay cubes – Combination of Delta and full refreshes – Expect more deltas as time becomes an issue 22/09/2007 11 We had the Politics Business case from hell Executive turnover – A champion, a champion – my kingdom for a champion Budget blues – Project drawn up & commenced in a year of SARS SARS--induced reversal on international income Business Case RFI ITO Delays due to (un)due process and staff unavailability Manyy selfself-appointed pp ((and opinionated) p ) Data Warehouse Experts 22/09/2007 12 Expect Complications Everyone wants to dump their old data in the warehouse – it’s a WAREHOUSE not an OUTHOUSE!!!! ETL tools are powerful, complex and challenging beasts – especially when in unfamiliar territory (but that’s part of the fun) Executive turnover + University review + Reorganisation + … meant Ever moving interview target 22/09/2007 13 Golden (Midas) Moments First query on GPA at Melbourne Campus on prepre-release Unfortunate clarity on drop in student numbers Look on various faces when they see data on maps with drillthrough But for every Midas moment there’ll there ll be months of mud – Strangely g y many y occur in lead up p to Project j Board meetings ???? 22/09/2007 14 What have we achieved so far? Student Operational Data Store Student Data Marts – – – – Student Course Student Program Student Applications Student Prospects HR ODS now available – Meta Layer plus other tables – Security of access based on org structure security extracted from Alesco application Other data feeds – AIC Staff St ff portal t l – Faculty restructure – Axapta (used to manage external student mailouts and packages) 22/09/2007 15 Current CQU IS Environment PS Financials 8 8.4 4 Upgrading student from 7.6 to 8.9 HR system t (not ( t PS) upgraded d d tto web b based architecture November 06 Cognos 7 reporting tools – at least the majority we’ve deployed – are not So we had to find a new mechanism for the many y reports p we have available 22/09/2007 16 So in June 2006 Decision to look for web based reporting tools Selected Cognos 8 after doing research research, leveraging existing work with some migration aspects covered, covered and some not Training in late October 06 22/09/2007 17 Approach taken Pause work in warehouse due to resource limitations A bit of training never hurts, neither leveraging off a pretty useful consultant Migrate Ye Olde Ancient Application infrastructure and their reports first – A lowlow-risk familiarityy with the new technology gy – Doesn’t hurt to do a few simpler things first – Legacy g y student and finance data complete 22/09/2007 18 There will always be gremlins… Impromptu p p catalogs g do migrate g to COGNOS 8 framework packages… … but … Migrations do not in and of themselves make industrial strength infrastructure … … particularly ti l l if th there’s ’ a very significant i ifi t paradigm shift Migrated infrastructure a starting point No ROI on “tarting up” legacy system Impromptu catalogs g 22/09/2007 19 Sometimes Sometimes, serendipity serendipity… Cognos 8 Framework Manager is very clever in dealing with star schemas Designate g dimensions,, fact tables,, and scope p Auto aggregation and other good gear follows … which only goes to show Star Schemas make infrastructure as well as reporting easy … which c makes a es Ken e a very e y happy appy ca camper pe 22/09/2007 20 New Boss Arrives Emphasis p to shift to exec interface - dashboards and high value adds adds “Strategic” is my mantra BUT – Need to convert existing report suite to COGNOS 8 as well – New reports for HR & Student Upgrade – Staff resourcing issues Trips to two other unis in Oz to have a look and share ideas, problems and solutions 22/09/2007 21 Reporting Pyramid We had successfully used our reporting tools in the operational and management areas with some “bleed” areas, bleed into strategic Not helping executives who need information “no no further than a click away” away and exposure to “games” some sections of the University play Dashboarding now a major focus 22/09/2007 22 The Fatal Flaws (and CQU’s Score on some off them)) COGNOS has a white paper on these Available from their web site B aware that Be th t the th 7 d deadly dl d dashboard hb d don’ts (TM Ken D 2007) will vary according di tto which hi h company produce d them Briefly cover each and how well CQU has done / is going 22/09/2007 23 1 1. “If If we build itit, they will come” come Gartner says y a successful initiative combines business relevance with strong architecture Ours works depending on a favourable definition off “they” “h ” Having the data is part of it but it will remain untapped without buy buy--in and appropriate delivery mechanisms Change C a ge & co communication u cat o management a age e t is s key ey CQU rating (/10): 3 In summary, y, g guilty y as charged g 22/09/2007 24 2. “Managers need to dance with the numbers”” Spreadsheet proliferation – everybody has their own version of the numbers Cf “One Cf. One version of the truth” truth Very important for data to be presented f free off nuance and d misinterpretation, misinterpretation i i t t ti , and secure from “creative flair” CQU Score – N/A. Yet. Turf war may be approaching 22/09/2007 25 3. “We don’t have a data quality problem”” We know _most_ of the student data issues Errors reported to users for correction at source Error dimension rows created so that data is visible where valid and identified where not We are becoming painfully aware of HR issues Finance will be “interesting” Data cleansing major byby-product of exposing “information” Diefenbach’s Doctrines of Data Daintiness – Data will appear to all intents and purposes clean and clear until you use it – Do not for one moment consider you will find all the holes first – Nothing “airs one’s dirty linen” or “crap detects” like 3rd party tools CQU score 66-7/10 22/09/2007 26 4. “Our business application vendor will deliver the best solution”” Not going there Have high opinions of and more than significant relationships with both COGNOS and Oracle and they’ll give you different interpretations Remember, it has to work for your I tit ti Institution 22/09/2007 27 5. “Darwin was wrong. There’s no need ffor BI applications to evolve”” In which COGNOS p pushes a single g tool set to cover all BI Certainly makes maintenance easier, and from experience i we kknow “f “feral” l” BI apps can breed b d like rabbits without the rigour & governance that a single toolset can provide Points of pain particularly occurs when one exec discovers data delivered differently to another exec and it’s data that they want as well CQU 6/10 22/09/2007 28 6. “We can outsource the whole thing”” Depends on how you define “the the whole thing” Not touching this one in print either ☺ 22/09/2007 29 7 7. “Just Just give me a dashboard” dashboard Dashboards often win over balanced scorecard because they’re seen to be easier and faster to deliver M Must reflect fl b business i d drivers i and d allow ll d drill drillilldown to supporting detail More effective if based on strategy maps which show cause and effect Easier as e to do with t a warehouse a e ouse u underpinning de p g And we haven’t done enough in this space for me to comment 22/09/2007 30 Diving in for a Closer Look 22/09/2007 31 Stars in Our Eyes We chose Kimball Star Schema approach as it was our preferred of 2 approaches available cff Bill Inmon’s approach I Geekspeek: G k k In – (Conformed) Dimensional Tables shared across all Fact Tables 22/09/2007 32 St d t Program Student P Data D t Mart M t – Every E Student Program g (since ( 1989 anyhow) y ) – – – – – – Total Courses Registered Total Courses Dropped T t l Courses Total C E Enrolled ll d Total Courses Passed Total Courses Failed Total Courses Outstanding – Total Credit Points Attempted – Total Credit Points Passed – Total Credit Points Outstanding 22/09/2007 – Weighted GPA – Total Grade Points – Total T l GP Credits C di T Taken k – Total Workload Attempted – Total Workload Passed (EFTSL) – Total Workload Outstanding (EFTSL) – CQU Transfer Credits – Other Transfer Credits – Total T t l Transfer T f Credits C dit 33 Student Program Dimensions Student Personal Details – Personal P l iinformation f ti about b t each h student, t d t iincluding l di names and d addresses, biodemographic data, citizenship and presence at the University. Program Details – Information about each academic program. Latest Student Program Details – The latest information about each student program. Expected Final Term Details – The term the student expects to finish the program. Completion Term Details – The actual term the student completed p the p program. g First Student Program Details – The earliest information about each student program. Application Details – Some S information i f ti about b t th the student t d t program application. li ti 22/09/2007 34 Latest Stuff / Coming (Very) S Soon… oon… Access by new Faculty Structure – SCD’s SCD’ weren’t ’t meantt tto cope with ith “Bi “Big Bang” B ” Which teaching week every course was added or dropped during semester – from 6 weeks or more prior to Week 1 to 10 or more weeks into term – NB vacations not considered teaching weeks OP Score / Rank on entry but it isn’t quite where it should be yet Cohort Sequencing for Dummies – – – – – – Nth program commenced db by student t d t M awards from this program This is the Pth award received This student has commenced X p programs g overall This student has Y awards across all programs Probably by early 08 Any significant requests 22/09/2007 35 Gut Clenching Finger Crossing Demonstration Time Remember Diefenbach’s Diefenbach s 2 axioms of demonstrations: – “The The chances of a demonstration working is inversely proportional to the square of the organisational g p power and influence of the audience” – “The typing yp g mistake rate increases proportionally to (n+1)2 where n is the number of observers” 22/09/2007 36 Questions? 22/09/2007 37