A forecast that is right today is wrong in 4 weeks. The world moves: comp-set rates change, new flight routes open, a competitor renovates, your own brand pivots. The discipline that separates a 70%-accurate forecaster from a 95%-accurate one is the recalibration cycle — the standing routine where you compare predictions to actuals and update the model.
Step 1: lock the prediction
At the start of each 4-week window, write down your forecast for the 90-day horizon. Specifically: total roomnights, ADR, RevPAR by week. Save it. The saved version is the prediction you will be measured against. Without locking, you will unconsciously revise your forecast as actuals come in — which is the opposite of forecast discipline.
Step 2: track the actuals weekly
Each Friday, plot actual roomnights, ADR, and RevPAR against the locked forecast. Calculate absolute variance per week and cumulative variance for the period. A property where weekly variance stays under ±5% has a tight forecast. A property where weekly variance is ±15% is operating on guesswork.
Step 3: identify what broke
When variance exceeds the threshold, the question is not "by how much" but "why." Three common drivers:
Step 4: update the model
At the end of the 4-week cycle, you have either a tight forecast (no model change needed) or evidence of specific bias (e.g., "we consistently underestimate German wholesale by 8-12% in week 22-30"). Adjust the assumptions in the forecast model, document the change, and lock the next 4-week prediction.
After 6-8 cycles of this discipline, a property's forecast accuracy reliably sits in the ±3-5% band. That is the level at which ownership stops asking second-guessing questions in the monthly review.