A/B Player Explorer
Who drives the lift, and who leaves
Spend per paying user (ARPPU) is … in B versus A, while paying users are … and conversion is …. The lift comes from existing payers spending more, not from more payers.
A is the control arm, B the treatment. Paying users are users with at least one purchase. Lifts are relative, B versus A, over the whole test.
Payer spend distribution
Total revenue per paying user. Each bar is the share of that arm's payers in the bin. Pick a percentile to mark it on the chart.
Revenue concentration
Share of payer revenue brought by the top payers of each arm. The dashed diagonal is perfect equality.
Active days: who leaves
Disengaged means N or fewer active days. Move N to see how the share of disengaged users compares between arms.
The gap peaks at N = … (…). By N = … the cumulative shares are … in A and … in B, so B removes nobody. It moves the same segment (… users in A, … in B) about … earlier: … active days on average in A against … in B.
Spend vs engagement
Mean revenue per paying user by number of active days. Hover a group for the arm gap.
Results by active-days bucket
The same figures as the chart, with each bucket's share of its arm's payer revenue.
| Active days | Arm | Payers | Mean revenue | Revenue share |
|---|
Source: A/B test of a feature rollout in a mobile game, repository proggdd/ab-test-mobile-game. Payers reach at most … active days in this data, so higher buckets are empty and left out.