Forecasting attendance from weather, holidays, and the calendar

Attendance drives revenue, labor, and purchasing, but it had been forecast largely from seasonal averages. Weather, holidays, and calendar shifts moved actual attendance far enough from the plan to affect staffing and the revenue outlook.

An attendance forecast that cannot explain itself cannot be challenged or improved. Leaders need to know whether a variance came from the weather, the calendar, or the business.

• Aligned prior-year days by day of week and holiday, not calendar date, so that comparisons are like-for-like.
• Modeled the effect of weather and holidays on attendance for each venue separately, since the same driver can affect different venues differently.
• Tested each driver for a real effect before including it — and left out drivers that could not be separated from ordinary seasonality.
• Connected the forecast to the budgeting platform so labor and revenue plans use the same attendance assumption.

• A driver-based daily attendance forecast with weather and holiday adjustments by venue.
• Variance explanations that attribute differences from plan to weather, calendar, and underlying demand.
• One attendance assumption shared by revenue, labor, and operating plans.

• Forecasts that explain their own movements, so leadership can act on the drivers.
• Labor and revenue planning working from the same attendance number.
• A disciplined bar for which demand drivers are allowed into the plan.

• Regression & seasonality analysis
• Weather and holiday effects
• Prior-year date alignment
• Driver testing
• Python
• SQL