3 common mistakes when looking at freemium metrics
DESCRIPTION
Stam Beremski @ NaturalMotion presentation for 2013 Games Industry Analytics ForumTRANSCRIPT
3 COMMON MISTAKES WHEN LOOKING AT FREEMIUM
METRICS
Sunday, 3 February 13
INTRODUCTION
• Product Lead at NaturalMotion
• Previously worked on:
Sunday, 3 February 13
OVERVIEW OF FREEMIUM GAMES
Sunday, 3 February 13
THE FREEMIUM MODEL
App Store
Engage
Retain
Monetize + Invite
Sunday, 3 February 13
Good metrics give you insight into conversion at each stage of the model
App Store
Engage
Retain
Monetise + Invite
Good metrics have explanatory powers
Sunday, 3 February 13
MISTAKE #1
NOT FOCUSING ON EXPLANATORY METRICS
Sunday, 3 February 13
• Total Revenue / Total Users• Session per player / day• Invites sent / player• Invites accepted / Invite• ARPDAU• DAU/MAU
Vanity Explanatory
• Users• Page Views• Daily Revenue• Total Mins of Play• Total Sessions
• Cohort segmentation• Retention• LTV• Sessions pp / day
Counts Ratios
• Behavioural segmentation• Whales vs Free• Single player vs Multi
• Funnels• Tutorial / Quest• Virality funnel
Segmented Ratios
Sunday, 3 February 13
MISTAKE #2
NOT CONTROLLING FOR VARIABLES
Sunday, 3 February 13
You have just released an update to your game and you take a look at the metrics to gauge success...
• Did this update improve monetisation?
• How long did it take to us get a conclusive answer?
(Data is fictional)
Sunday, 3 February 13
There is a problem if you only look at New Users, Revenue and ARPDAU to gauge the success of an update...
Solution: Look at metrics which isolate what you are interested in and control for other variables (in this case the player’s lifecycle within the app)
Problem: Metrics can be affected by uncontrolled variables• In our fictional game players spend 90% of their LTV within 7 days of first playing the app
• An large influx of new players will cause a revenue spike even if the app remains unchanged
(Data is fictional)
Sunday, 3 February 13
The solution is to look at player LTV at specific points in the player’s lifecycle
LTV = Total Revenue from cohort
Total # Player of a cohortDay N LTV = Player LTV N days into
the lifetime of a cohort
(Data is fictional)
Sunday, 3 February 13
MISTAKE #3
NOT LOOKING AT THE DISTRIBUTION OF
UNDERLYING DATA
Sunday, 3 February 13
We often look at averaged data.
Mean = 3.89Median = 3Mode = 3
{1,2,3,3,3,4,4,5,10}
Sunday, 3 February 13
Power-law distribution are common in freemium games
Max: $624Min: $0Mean: $0.57Median: $0Mode: $0Std Dev: 7.12% Payers: 5%
Represents $65,000 revenue from110,000 players
(Data is fictional)
Sunday, 3 February 13
But statistics can be misleading...
Property Value
Mean of x in each case 9 (exact)Variance of x in each case 11 (exact)Mean of y in each case 7.50 (to 2 decimal places)Variance of y in each case 4.122 or 4.127 (to 3 decimal places)CorrelaAon between x and y in each case 0.816 (to 3 decimal places)Linear regression line in each case y = 3.00 + 0.500x (to 2 and 3 decimal places, respecAvely)
Sunday, 3 February 13