A monetization target on its own is incomplete. Write down what the additional revenue is allowed to cost — before the results arrive.
Thresholds vary by game, cohort, stage and business model. What matters is deciding them before the results arrive.
Target: metric + expected lift + read dateGuardrail: metric + maximum acceptable deteriorationStop rule: condition that triggers rollback“Raise ad ARPDAU 10% via interstitial frequency; protect D7 retention (max −1.5 pts) and payer conversion (max −5% relative); roll back if either breaches at the D14 read.”
Defaults on this page are illustrative starting points, not benchmarks for your game. Replace them with your own numbers.
A metric the experiment is not allowed to damage beyond an agreed threshold while it optimises its primary metric. For a monetization test, typical guardrails are retention, payer conversion and session depth.
Revenue moves quickly; retention, payer behaviour and paid-cohort payback move slowly. If the acceptable trade-off is decided after the results arrive, the team judges it knowing which metric improved — which biases the call.
A specific, observable condition and a date: “roll back if D7 retention in the test group falls more than 1.5 points below control at the D14 read, or payer conversion falls more than 5% relative.” Vague rules get renegotiated mid-test.
The change in evidence or context that would reopen the decision after rollout — for example, a later retention decline or a shift in traffic mix. It stops temporary evidence becoming a permanent rule.