Space Ape's Analytics Stack
22 likes10,954 views
The document outlines the development and analytics of the mobile game 'Samurai Siege', which launched in October 2013 and achieved 11 million installs with $23 million in revenue. It discusses the data handling and analytics processes, including user acquisition tools, CRM tools, and automated reporting to derive insights from player behavior and engagement. The necessity for a unified data dashboard and predictive analytics to improve user retention and monetization strategies is emphasized.
Space Ape's Analytics Stack
- 2. 1. Samurai Siege Overview 2. Samurai Analytics Stack Evolution 3. DWH Components 4. User Acquisition Tools 5. Product Tools 6. CRM Tools
- 4. Mobile real time strategy, iOS, Android Developed in 6 months by team of 12 Live October 2013 11M Installs / 1.2M MAU / 250K DAU $23M in Gross Revenue to Date
- 5. Small dev teams Creative + Science Agile Live Operations Data empowered
- 7. 135 Billion Rows 13TB Of Compressed Data 250m x 125 columns Rows Per Day
- 8. DATA MINING ANALYSIS REPORTING CRM Data Capture A/B Tests Samurai Siege Data Stack Disconnected from live game Insights & Learnings Weekly reporting ETL
- 11. ● Automated and Run Daily through an SQL Runner ● Delta Process. Clear and Insert Data for Past 3 Days ● Process completed in Minutes not Hours
- 13. Platform Spend BehaviourDevice Retention Country Language Acquisition Channel Game Balances Operating System Castle Level Retention
- 14. player_id 71724ca2-c1f8-4519-8ae5-f8628a005653 villagename XXXXXXXX allianceid 9b9f3aaa-09ee-42f9-b53f-5f9003e59c79 alliancename XXXXXXXX signupdate 17/08/2013 06:39:00 max_actiondate 08/02/2015 17:58:00 controlflag 0 playerlevel 35 castlelevel 9 max_pveseq 11000 playerhonor 411 cohort 2 tenure 540
- 15. first_deviceraw iPad2,1 first_devicelong iPad2 - Wi-Fi first_deviceshort iPad2 first_playerlanguage English first_deviceos ios first_deviceplatform ios first_deviceid c82848be88d034317526fa72b1b597ad last_deviceraw samsung SM-G900F last_devicelong Samsung Galaxy S5 (Europe) last_deviceshort Samsung Galaxy S5 last_playerlanguage English last_deviceos android last_deviceplatform android
- 16. first_country_iso US last_country_iso GB first_country UNITED STATES last_country UNITED KINGDOM first_continent NORTH AMERICA last_continent EUROPE email XXXXXXXXXXX ret_d1 0 ret_d3 0 ret_d7 0 ret_d12 0
- 17. spend_first24hours $ - spend_first3days $ 9.00 spend_first7days $ 9.00 spend_first2weeks $ 9.00 spend_last24hrs $ - spend_last0to7days $ - spend_last8to14days $ - spend_last15to21days $ - spend_last22to28days $ - spend_total $ 104.00 last_purchasedate 18/04/2014 00:00:00
- 18. last_purchasedate 18/04/2014 00:00:00 tot_login_first7days 86 login_last0to7days 0 login_last8to14days 0 login_last15to21days 1 login_last22to28days 1 spendtier Villager is_newbie 0 is_nonspenderprio 0 is_spendrisk 0 publishername ORGANIC campaignname subadgroup publisherid
- 19. ● Automated and Run Daily through an SQL Runner ● Recreated in Full every day ● Based on Daily Summary Table ● Process completed in Minutes not Hours DAILY SUMMARY
- 23. Install Date Campaign Network Campaign Name Country Cost Clicks Impressions
- 25. SEGMENTATION BASED ON MONETISATION AND ENGAGEMENT IN THE FIRST 7 DAYS
- 27. ● Segment Users ● Assign Appropriate Curve based on Segmentation ● Calculate pLTV based on different time periods ● Automated and Run Daily through an SQL Runner
- 28. Fragmented Area Data Driven Unified Dashboard is needed to scale Fast Reaction
- 32. Retention Average Revenue Per User Average Revenue Per Paying User Daily/Monthly Active Users New Users Session Length Average Sessions Per User Spend Metrics
- 33. Spender Conversion % Life Time Value Re-engagement
- 34. Spend Tiers Time Cohorts Acquisition Channel Device Platform Country In Game Segments e.g. Castle Level
- 36. Set up A/B test in SWRVE + basic results More advanced metrics analysis in DWH
- 37. Army composition for castle 9 and 10 ? in Japan, during a event, ...More dimensions, aggregation flexibility ● Monthly / weekly revenue ● Full device split ● Event type ● ... To be applied on a lot of information ● IAPs , diamonds spending ● Economy movements ● Event performance ● Honor movement ● Behavioural changes (eg army composition) ● Community stats (time to first response, CSAT, per value tier) ● ...
- 39. BEHAVIOURAL SEGMENTS SPEND LEVELS EMAIL MESSAGE DETAILS
- 48. Decide from the Beginning if you want to create a Data Warehouse
- 49. Do not collect everything!
- 51. Keep your finger on the pulse
- 52. Automate!
- 54. Using Predictive Analytics to Trigger In- App Messages/Offers for users.
- 55. Enhancing our Machine Learning capabilities
- 56. Network / Cross Platform View and Analysis