ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

Game Analytics: A Practitioner¡¯s Perspective

?Download as PPTX, PDF?
?
D
Decimus

Presentation at IEEE Game Innovation Conference - 23 Dec 2010 http://ice-gic.ieee-cesoc.org/2010/index.php

1 of 38
Downloaded 36 times
Game Analytics:A Practitioner¡¯s Perspective2nd IEEE Consumer Electronics Society Games Innovation Conference23 December 2010
Game Analytics is crucial to the successof any online game
Peter BurtonCivil Engineering first degreeMBAWorked mainly in:Retail Banking and Insurance IndustriesCall centre consultancyMobile TelcosOnline advertisingMost recently applying lessons learnt from beinga ¡°white coated¡± media accountant to the online gaming industry
OutlineOutblazeAdserving and Game Analytics (GA)Learning from other industriesA crash course in GA jargonSome analysis toolsSummary
Privately held company founded in Hong Kong in 1998 with staff in Hong Kong, US, UK, Korea, Singapore, Philippines, MalaysiaOriginally a pioneer in white label messaging, managing 40+m active mailboxes per month. Outblazesold its messaging business to IBM in 2009New focus is: Entertainment and MediaVideo GamesSocial, Community and Online Multiplayer Games
Online GamesCommunity Web SitesHello Kitty OnlineFlash GamesBen10DS & WiiGamesAnimatedTV showsPublishingPC GamesIn APACVirtual RealityProductsiPhone AppsHello KittyFacebook Apps
Adserving
Data RULES
Game companies are NO different fromAirlines, CC, Mobile Telcosor Supermarkets The data collected aboutuser activity providesdifferential competitive advantage
It¡¯s about making sense from unstructured dataIt¡¯s about knowing what and how to use that data It¡¯s about multi-disciplinary collaboration
Data is the ENGINE that helps you:Improve ROIMake better decisionsProvide a continuous feedback loop for your game designers, engineers, marketers game economists, ¡­Sometimes pinpoint unforeseen issues!
Some proxies for learningAirlines: like CXUpgrades linked to your direct Value to the airline.CRM.Aircraft provisioning.Yield Management.Demand based pricing.
Some proxies for learningSupermarkets: like TESCOWhere to place the biscuits?Edwina Dunn & Clive Humby of DunnHumby.Rolling Ball Analysis.
GA is Not:anymore about gut feelabout guessingnorIt would appear Hopkins that your gut feel was only indigestionIt¡¯s a guess. I never said it was an educated guess.GA helps REMOVE gut feel and guess workfrom the decision making process.
My KEY Audience at Outblaze is our:KaChing teamEngineersRetail mall teamsCustomer Service teamsGame developersMarketersGame economistPricing specialistmulti-disciplinary collaborationis KEY
The Jargon of GA
GAME TELEMETRYthe RAW data that is collected from online games
GAME TELEMETRYTelemetry is driven by event logs:name value pairs(which identify events)PLUS a unique session ID
GAME TELEMETRYBuild the event logsPLUSunique session IDs into the game at the game design stage
GAME TELEMETRYevent logsPLUSunique session IDsshould NOTbe written backto the Game dB!
GA means:analysing large volumes of in game data feeds to help us find, get & keep playersunderstanding player user behaviour so that we can continually improve our gamesplaying with the data to tease out trends and insights via what if scenariosproviding a near real time feedback loop for our designers, engineers and the game economists¡­
GA is about:making money using FACTS rather than gut feel or guesswork!understanding what satisfies players and making them come back, again and again and again (retention)understanding what drives the money, what works and what doesn¡¯t workmaking better games that generate better levels of return
GA is a mix of old school statistics combined with the latest in web engine metrics
GA typically usesmulti variable analysis:Players | Levels | Quests | Profiles |Dollars
Crash course in basic GA metricsActive Users (DAU, WAU, MAU) key measure DAU/MAUARPU (Average Revenue per User ¡­ over time)Churn or attritionCohortsEngagementEntry eventExit eventFunnelsK factor (virality)Lifetime valueRe-engagementRetentionSession analysisUser Acquisition Cost
Data SourcesGame logs | Mall | Payment gateway | Player profilePlayer insightsWhat they buy? | How they buy? | Why they buy?Game Designer Feedback LoopConnecting the dotsWho matters most? The 2% to 5%Driving engagement
Visualisation ImportantYour audience typically fall into 2 user categories:Your Numerate Audience(prefer a table of numbers)Your Visual Audience (the majority who prefer a pictureand pre-digested insights)
TOOLSIf you do use Excel you¡¯ll need Excel 2010 ¨C 64 bit with MAX RAM(plus the PowerPivot add in)Excel is not generally man enough!It¡¯s great for proto-typing
How to burn out an i7 laptop12G of RAM maxed out!All 12 i7 cores maxed out!The culprit; some excel magic which helps to calcuniquesby day and by week and by year:=IF(SUMPRODUCT(($A$2:$A500,000=A500,000)*($S$2:$S500000=S500,000))>1,0,1)Col A is the user-id, Col S is the day flag.?The output of the above is a count of the unique user-ids for a given day.For proto-typing, I now use a desktop i7 with 12 cores, 24G of RAM with 8 fans! It¡¯s a MONSTER gaming machine without a superadoopergraphics card!
What we use¡­a data warehousing and data visualisation tool to automate data feeds and provide a suite of standardised reports¡­ plus the tools to allow us to ¡°dive¡± into the raw data and slice and dice it
Game Analytics: A Practitioner¡¯s Perspective
Game Analytics: A Practitioner¡¯s Perspective
Game Analytics: A Practitioner¡¯s Perspective
Game Analytics: A Practitioner¡¯s Perspective
Game Analytics: A Practitioner¡¯s Perspective
Game Analytics: A Practitioner¡¯s Perspective
Summary ¨C 6 keysGA is crucial for any game. It will help you:find, get and keep players focuses effort on ROI (2%-5%)provide a near real-time feedback loop for game designers, engineers, marketers and game economists¡­to better understand player behaviour and open the way for continuous game improvements
Summary ¨C 6 keysTELEMETRY is key.Build it in at game design stageEvent flags (name value pairs) need to be combined with a unique session ID DO NOT write Telemetry back to the dB!VisualisationMD collaboration / communicationEmploy smart gradsLearn from others within and outside of your industry

More Related Content

Game Analytics: A Practitioner¡¯s Perspective

  • 1. Game Analytics:A Practitioner¡¯s Perspective2nd IEEE Consumer Electronics Society Games Innovation Conference23 December 2010
  • 2. Game Analytics is crucial to the successof any online game
  • 3. Peter BurtonCivil Engineering first degreeMBAWorked mainly in:Retail Banking and Insurance IndustriesCall centre consultancyMobile TelcosOnline advertisingMost recently applying lessons learnt from beinga ¡°white coated¡± media accountant to the online gaming industry
  • 4. OutlineOutblazeAdserving and Game Analytics (GA)Learning from other industriesA crash course in GA jargonSome analysis toolsSummary
  • 5. Privately held company founded in Hong Kong in 1998 with staff in Hong Kong, US, UK, Korea, Singapore, Philippines, MalaysiaOriginally a pioneer in white label messaging, managing 40+m active mailboxes per month. Outblazesold its messaging business to IBM in 2009New focus is: Entertainment and MediaVideo GamesSocial, Community and Online Multiplayer Games
  • 6. Online GamesCommunity Web SitesHello Kitty OnlineFlash GamesBen10DS & WiiGamesAnimatedTV showsPublishingPC GamesIn APACVirtual RealityProductsiPhone AppsHello KittyFacebook Apps
  • 9. Game companies are NO different fromAirlines, CC, Mobile Telcosor Supermarkets The data collected aboutuser activity providesdifferential competitive advantage
  • 10. It¡¯s about making sense from unstructured dataIt¡¯s about knowing what and how to use that data It¡¯s about multi-disciplinary collaboration
  • 11. Data is the ENGINE that helps you:Improve ROIMake better decisionsProvide a continuous feedback loop for your game designers, engineers, marketers game economists, ¡­Sometimes pinpoint unforeseen issues!
  • 12. Some proxies for learningAirlines: like CXUpgrades linked to your direct Value to the airline.CRM.Aircraft provisioning.Yield Management.Demand based pricing.
  • 13. Some proxies for learningSupermarkets: like TESCOWhere to place the biscuits?Edwina Dunn & Clive Humby of DunnHumby.Rolling Ball Analysis.
  • 14. GA is Not:anymore about gut feelabout guessingnorIt would appear Hopkins that your gut feel was only indigestionIt¡¯s a guess. I never said it was an educated guess.GA helps REMOVE gut feel and guess workfrom the decision making process.
  • 15. My KEY Audience at Outblaze is our:KaChing teamEngineersRetail mall teamsCustomer Service teamsGame developersMarketersGame economistPricing specialistmulti-disciplinary collaborationis KEY
  • 17. GAME TELEMETRYthe RAW data that is collected from online games
  • 18. GAME TELEMETRYTelemetry is driven by event logs:name value pairs(which identify events)PLUS a unique session ID
  • 19. GAME TELEMETRYBuild the event logsPLUSunique session IDs into the game at the game design stage
  • 20. GAME TELEMETRYevent logsPLUSunique session IDsshould NOTbe written backto the Game dB!
  • 21. GA means:analysing large volumes of in game data feeds to help us find, get & keep playersunderstanding player user behaviour so that we can continually improve our gamesplaying with the data to tease out trends and insights via what if scenariosproviding a near real time feedback loop for our designers, engineers and the game economists¡­
  • 22. GA is about:making money using FACTS rather than gut feel or guesswork!understanding what satisfies players and making them come back, again and again and again (retention)understanding what drives the money, what works and what doesn¡¯t workmaking better games that generate better levels of return
  • 23. GA is a mix of old school statistics combined with the latest in web engine metrics
  • 24. GA typically usesmulti variable analysis:Players | Levels | Quests | Profiles |Dollars
  • 25. Crash course in basic GA metricsActive Users (DAU, WAU, MAU) key measure DAU/MAUARPU (Average Revenue per User ¡­ over time)Churn or attritionCohortsEngagementEntry eventExit eventFunnelsK factor (virality)Lifetime valueRe-engagementRetentionSession analysisUser Acquisition Cost
  • 26. Data SourcesGame logs | Mall | Payment gateway | Player profilePlayer insightsWhat they buy? | How they buy? | Why they buy?Game Designer Feedback LoopConnecting the dotsWho matters most? The 2% to 5%Driving engagement
  • 27. Visualisation ImportantYour audience typically fall into 2 user categories:Your Numerate Audience(prefer a table of numbers)Your Visual Audience (the majority who prefer a pictureand pre-digested insights)
  • 28. TOOLSIf you do use Excel you¡¯ll need Excel 2010 ¨C 64 bit with MAX RAM(plus the PowerPivot add in)Excel is not generally man enough!It¡¯s great for proto-typing
  • 29. How to burn out an i7 laptop12G of RAM maxed out!All 12 i7 cores maxed out!The culprit; some excel magic which helps to calcuniquesby day and by week and by year:=IF(SUMPRODUCT(($A$2:$A500,000=A500,000)*($S$2:$S500000=S500,000))>1,0,1)Col A is the user-id, Col S is the day flag.?The output of the above is a count of the unique user-ids for a given day.For proto-typing, I now use a desktop i7 with 12 cores, 24G of RAM with 8 fans! It¡¯s a MONSTER gaming machine without a superadoopergraphics card!
  • 30. What we use¡­a data warehousing and data visualisation tool to automate data feeds and provide a suite of standardised reports¡­ plus the tools to allow us to ¡°dive¡± into the raw data and slice and dice it
  • 37. Summary ¨C 6 keysGA is crucial for any game. It will help you:find, get and keep players focuses effort on ROI (2%-5%)provide a near real-time feedback loop for game designers, engineers, marketers and game economists¡­to better understand player behaviour and open the way for continuous game improvements
  • 38. Summary ¨C 6 keysTELEMETRY is key.Build it in at game design stageEvent flags (name value pairs) need to be combined with a unique session ID DO NOT write Telemetry back to the dB!VisualisationMD collaboration / communicationEmploy smart gradsLearn from others within and outside of your industry

Editor's Notes

  • #32: 1. Integ -?use drag and drop and link to create logic flow of data?- Filein : read content from local source file?- Calc : calculate intermediate variables, format variable, or if-then-else?- Sort : sort data in any dimension of your data?- Lookup : join two files?- Fileout : output content into file2. Builder?- decide how to look?into the model?and the statistics you are interested?- Dimension : the field you want to jump into?- Summary : the field you are interested in