Demand forecasting, anomaly detection, and crowd-shaping interventions for the 56 most-trafficked US theme parks. We turn signals into the three actions your duty manager should take in the next 90 minutes.
| Type | Sent | Accepted | Avg Lift |
|---|---|---|---|
| Zone Auction | 412 | 71% | +18% zone activity |
| Closing Push | 188 | 74% | −22% close-time stacking |
| Chain | 224 | 56% | +1.8 rides/guest |
| Flash | 612 | 41% | +14% credit yield |
| Credit Proximity | 188 | 79% | +0.9 cards/guest |
| Early Bird | 96 | 68% | +12% morning rides |
90.3% within 1 decile, ±8 min wait MAE — measured by leave-one-park-out CV across 52 parks. Accuracy holds at T-14 days because calendar and seasonal signals dominate the model.
Nine behavioral-incentive types (zone auction, closing-hour push, chain, flash, credit proximity) deployed against the loss-aversion, endowed-progress, and variable-reward literature. Operator-tuned, cohort-routed.
±8 min wait forecast · 90.3% crowd accuracy · 5 min detection latency. Hit these in 60 days or you don't pay for the pilot. Built on 20,360 park-days and counting.
30 minutes. We pull a real day from the past 30, walk through 4 moments where Thoosie would have changed a decision, then talk pilot scope.