The question a bonus report should answer is not how many bonuses were issued, but whether the players who received them stayed. That is a cohort question, and cohort questions are easy to get subtly wrong: an unbalanced comparison, a control that is not comparable, or a blended average that hides the tier where the money actually went. This guide sets out how to measure bonus impact so a CRM lead can take a defensible position to finance rather than a hopeful one.
Executive summary
Bonus impact is measured by following the players a promotion touched over the weeks after it, and comparing them with a group that behaved similarly but did not receive it. The measures that matter are the released cost (value that reached players after wagering, not the headline issued amount) set against the continuing contribution the cohort made. Reading those at the same cohort age, and by value tier rather than as one number, is what separates a promotion worth repeating from one that simply moved money out of the door. These are reference-design recommendations, not a description of an audited Bounty AI deployment.
Step 1: separate issued, released and contribution
Three numbers describe a bonus, and conflating them is the most common analytical error. Issued is the headline. Released is what actually reached players after wagering requirements played out. Contribution is what those players went on to generate afterwards. A large issued figure tells you the size of the gesture; only released cost against later contribution tells you the result.
Establishing these three cleanly, on one set of definitions, is the precondition for everything that follows. A cohort built on issued amounts will overstate cost; one that ignores wagering will misjudge which offers were efficient. The remaining steps, building the treated cohort and its control, reading the post-bonus ramp, and breaking the result down by value tier, all depend on getting these three quantities and their timing right first. Each step below is shown with the product screens for the view it describes.