Value-Weighted Retention Targeting Can Concentrate Harm

Retention incentives in online gambling are often allocated by ranking players on estimated treatment effect multiplied by lifetime value. Under a fixed budget that rule maximises operator value, but operator value and player harm can rise together.

The paper shows that the harm exposure of any budget-exhausting targeting rule exceeds the population mean by the covariance between treatment and harm, divided by the budget. It then gives sufficient conditions for value-weighted ranking to make that gap non-negative.

Main findings

• In a simulation where harm rises with value, value-weighted targeting raises mean harm among treated players from 0.421 to 0.476 at a 30% budget, and further at a 10% budget.
• Harm tracking value is not enough on its own: in one design an accurate value-weighted rule falls below the population mean despite a strong harm-value correlation.
• A soft harm weight on the score brings exposure down to roughly what effect-only ranking reaches with no harm input at all.
• A hard harm cap only beats the soft weight when the model’s components are accurate, and its ceiling holds only for the harm measure it is given.
• With estimated effects, ranking on the effect alone beats value weighting for every learner at budgets up to 30%.

All parameters are illustrative. The whole study is deterministic: every number in the paper is generated from the code, and the published result files carry checksums so a rerun can be verified.