The personalisation signals that actually move retention
Game and content discovery is a retention problem before it is a recommendation problem. What to model, what to ignore, and how to prove uplift.
Whizdom AI · Product team · 19 March 2026 · 6 min read
A player who cannot find something to play does not complain. They leave. That makes discovery one of the few retention levers that never appears in a support queue and rarely appears in a churn survey.
Behaviour beats demographics
Interaction with the product — what a player opens, returns to, abandons and plays at which time of day — predicts the next session far better than any segment definition written in a workshop. Recommendations should respond to in-session behaviour, not to yesterday's batch export.
One model family, both verticals
Sportsbook and casino are usually personalised by separate tools that never share a view of the player. Treating them as one surface means a player's casino behaviour can inform their sportsbook lobby, and vice versa.
Prove it with a control group
Every rollout should run against a control group so uplift in turnover per player, bets per player and retention is evidenced rather than asserted. If a personalisation vendor cannot show you the control arm, they are showing you a trend line.
Display relevant sports betting and casino content based on the player's interaction with your product.