This document discusses how mobile game developers can take a data-driven approach to optimizing the user experience on iOS and Android platforms. It covers measuring key metrics like retention, monetization, engagement, and virality to understand player behavior. Developers can then segment players based on patterns in the data and tailor the game experience for different segments. The goal is to continuously improve the game by being responsive to what metrics indicate players want.
Big Data in Mobile Gaming - Eric Seufert presentation from IGExpo Feb 1 2013
1. Big Data in Mobile Gaming:
Optimizing the User
Experience on iOS & Android
Eric Seufert
IGEXpo, Feb 1 2013
2. About Me
• Quantitative Marketer with focus on analytics
and statistical predictive methods
• MA in Applied Economics, BBA in Finance / CS
• Head of Marketing & User Acquisition at Grey
Area Labs (Helsinki)
– Previously strategy / analytics positions at Digital
Chocolate (Helsinki) and Skype (Tallinn)
3. About Me
• Blog frequently about F2P / “Big Data”
decision science / analytics at
http://ufert.se
• Feel free to reach out: eric@ufert.se
4. Presentation Agenda
• Introduction to “data-driven game design”
• Metrics portfolio – what’s important to track?
• Feature selection and analytics strategy
• Player segmentation and optimization
• Free-to-play design (spreadsheet included!)
• Q&A
5. What is “Big Data”?
• Buzzword, but with some substance
• Cheap storage = lots of data
• More data = less sophisticated statistical
techniques required to derive insights
• More data = easier to spot true trends (vs
anomalies)
6. What is “Big Data”?
• Example: A/B testing
• 200 data points = small sample size
– Does sample size reflect population? Must
perform t-test, possibly decrease confidence
interval
– What’s the distribution of data?
• 2 million data points = large sample size
– Distribution easier to infer
– Less testing / normalization required
7. What is “data-driven game design”?
• Iterative design
• Responsive to game metrics
• MVP -> behavioral feedback (metrics) ->
iterate -> update -> repeat
• Half intuition, half data science: fundamental
game mechanics established and then
tweaked based on player behavior
8. What is “data-driven game design”?
• Players get more of what they want
• Most enthusiastic players given latitude to
enjoy the game to greatest possible extent
9. The Metrics Portfolio
• Minimum set of metrics required to facilitate
data-based decisions
• Let players tell you what they like / don’t like
through their behavior
10. Four Key Metrics Groups
• Retention
• Monetization
• Engagement
• Virality
11. Retention
• D1-D7, D14, D28, D90, D365
• Most important set of metrics you track
• Communicates delight
– extent to which players enjoy your game = extent
to which they’re willing to return to it
12. Retention
• Retention “profile”: D1, D7, D28
• Follows general decay pattern:
– D1 *.5 = D7, D7 * .5 = D28
• Used to calculate player lifetime
• Look for D1 >= 40% for launch
• More info: http://bit.ly/WLvAOV
13. Monetization (F2P)
• Conversion rate: % of users that ever spend
money on IAPs
– 3-4% is good, >=5% high
• ARPU: Average Revenue per User
– Average amount of money users spend in-game
– Varies genre-to-genre, game-to-game
• ARPPU: Average Revenue per Paying User
– Greater the delta between ARPPU / ARPU, lower
the conversion rate
14. Monetization (F2P)
• ARPDAU: Average Revenue per Daily Active
User
– Average money spent per day by users
• Catalog distribution:
– Good to know where the bulk of revenue comes
from. Small purchases? Large purchases? How do
users engage with the product catalog? Best way
of visualizing is graphically.
15. Engagement
• Average Session Length / Median Session Length
– Daily lengths per user (what’s the average session
length?)
• Average Session Count / Median Session Count
– Daily counts per user (how frequently do they log in
each day)
• Medians important to track because very
enthusiastic players will skew the average (not
uncommon to see individuals logging in dozens of
times per day or playing hour-long sessions)
16. Virality
• K-Factor
– Number of users, on average, a single user
introduces to the game (>=1 = viral)
– Hard to track on mobile
– How to track k-factor when conversions are
unknowable (blog post: http://bit.ly/YB8BuL)
• You know purchased users, you can estimate organic
installs, back these out and lump the rest into “k-
factor”)
– Virality model: http://bit.ly/114n3gq
17. Analytics Strategy
• How do we know what to track?
– Retention is easy
– Monetization is easy
– Engagement is medium
– Virality is hard
• For the very basics, tracking should be
oriented toward first session and last session
• For the hard stuff (behavioral predictions),
tracking must be far more extensive
18. Two approaches
• I’m a smaller studio with less than 500k DAU. I
want to focus on the “big picture” stuff that
will help me retain players and boost
engagement
– Backend: standard SQL-based data warehouse,
nightly ETL
– Frontend: Dashboards communicate high-level
metrics across organization, PMs possibly use
some desktop visualization software to perform
basic analysis
19. Two approaches
• I’m a bigger studio with >500k DAU. I want to
focus on driving ARPPU and converting players
better
– Backend: Hadoop-based system to handle
processing large volumes of data
– Frontend: Dashboards + desktop analysis software
for PMs
– Analysis layer: team of analysts / data scientists
performing in-depth research, finding patterns,
and delivering insight to product teams
20. What’s the point of my analytics
system?
• Collect and process data that can be used to
derive insight that helps increase revenue
• Your analytics system should deliver $$$
• Analytics is not an intellectual exercise.
Analytics is the science of making more
money.
21. “Analytics” is vague. Who works in
analytics?
• The Analytics Engineer: system architect,
designs the tracking library and data transfer /
transformation / staging
• The Analyst: Handles reporting, dashboard
design, implementation
• The Data Scientist: Research and algorithmic
prediction. Helps drive data product design
(recommendation engines, matchmaking,
gameplay optimization)
22. What do I use analytics for?
• Optimizing the Player Experience:
• Give players the optimal experience based on
the information they’ve given you through
their behavior
• Improve the product catalog
– What do players want to buy? Where / when do
players want to buy? Are we accommodating this?
• UI / UX improvements through A/B Testing
25. My players are segmented. Now what?
• We have made some initial determinations
about this user based on his behavior
• Now we adapt the experience to fit this profile
• We measure exit metrics and iterate on
segmentation technique
26. F2P Design
• F2P is difficult
– Paradigm shift from paid downloads, not to
mention console titles
– Requires a different approach to design
• How do F2P games make money?
– LCV > CPA
– Continuous monetization curve
– High retention
– Low likely conversion
27. The F2P Model
• http://bit.ly/SlKYV0
• Projects revenues based on comprehensive
examination of user acquisition, virality,
monetization and retention
• Useful tool for making priority decisions on
game development
• Helpful when evaluating how much $ to invest
in user acquisition