Time series and forecasting · insight-0226

Model one-off events as interventions and keep them out of learned seasonality

Situation
A city hosted a single major sports final, and hotel and flight demand spiked for one week.
Action
Add a spike or short-window intervention dummy for the historical event so it does not contaminate other parameters. For a future one-off event, set the dummy only if you have analogous events, otherwise add a documented judgmental adjustment.
Reason
A spike variable absorbs a one-period effect like an outlier dummy, leaving trend and seasonal estimates clean.
Boundary
Concerts, strikes, outages, product recalls, system migrations.

Symptom or risk

Without an event indicator, the model treats the spike as seasonality or trend and repeats a smaller spike next year, or inflates its error variance.

Recurring events with stable timing (annual fairs) can be treated like holidays.

Judgmental uplifts for future events are hard to validate; record them so forecast value added can be measured later.

Adding a dummy for every awkward week until the model fits history perfectly.

Intervention variables include a spike variable for a one-period effect and a step variable for a permanent level shift.
Forecasting: Principles and Practice (3rd edition)