Foundation prediction schemas

One event per player, carrying their latest segmentations and the recommendations that follow from them. This is that contract — every field documented, with a live example.

Prediction event referenceThe full object: every field, type, enum and constraint.

Segment meaningsWhat each LTV, churn, incentive and reactivation value means.

What’s in an event

Lifetime value

Player spend tier, Low through VIP.

Churn risk

Likelihood of leaving — overall and per session.

Incentive usage & guidance

How a player uses bonuses, and what to offer next.

Reactivation potential

Odds of winning a lapsed player back.

Recommendations

Ranked incentives, favourite categories and games.

Player group

Target or Control, for measuring impact.

The shape

{
  "player": { "id": "PLAYER-123", "group": "Target", "metadata": { "siteId": 123 } },
  "customerLTV": {
    "current":  { "segment": "4. VIP",  "score": 1234.45, "sequence": 4 },
    "previous": { "segment": "3. High", "score": 980.10,  "sequence": 3 }
  },
  "recommendedGames": [
    { "rank": 1, "gameName": "Golden Whale Super Spinner", "score": 0.77 }
  ]
}

Two patterns run through every event:

• current / previous — each segmentation carries the latest prediction and the one before, so you can act on movement between segments, not just the current state.
• Target / Control — every player is tagged, so model-driven treatment stays measurable against a holdout.

See the full payload →