Динамическое ценообразование бонусов

Dynamic pricing бонусов — оптимизация промо-предложений под конкретного игрока для max ROI.

Задача

Каждому player: показать оффер с максимальной чистой прибылью. Trade-off: attractive enough для activation vs cost. Different segments отвечают на different offers. VIP не мотивирует €10 бонус. Casual не потянет €500 wagering.

ML подход

Uplift modeling: не just prediction, а effect of intervention. Two-model approach: treatment vs control. Meta-learners: T-learner, S-learner, X-learner. Contextual bandits для online learning. Features: LTV score, activity level, past bonus behavior.

Design constraints

Bonus abuse prevention: high risk users get lower bonuses. Legal constraints: same offer для same conditions в некоторых juris. Marketing budget: total spend cap. Regulatory: no predatory offers для problem gamblers.

Метрики

Incremental revenue: uplift vs no-offer control. Cost per activation. Post-bonus retention. Reactivation success rate. LTV growth. Compare vs static bonus strategies. Обычно 20–40% improvement possible от dynamic vs static.