A forecast is a decision aid, not a promise that the hotel will sell a particular number of rooms. Comparing it with what actually happened helps teams learn whether the assumptions, timing or available information need to change before the next planning cycle.
Forecast accuracy depends on clean definitions, stable data and the decision horizon. Review the Feathers Hospitality management approach, explore the operational platform, and talk with the team about performance reporting and planning.
At a glance
- Save forecasts with a date, horizon and assumptions.
- Compare like-for-like measures and explain material gaps.
- Use learning to improve decisions, not to rewrite the past.
Preserve the forecast as it was made
Store the forecast version, creation date, period covered, units, segments and assumptions. If the forecast is overwritten each week, the hotel cannot later tell what it expected at a given horizon or whether the update improved the decision.
Separate forecast rooms, room-nights, revenue and occupancy. These measures are related but not interchangeable. State how out-of-order rooms, complimentary stays, groups, cancellations and no-shows are treated.
Compare the same horizon and unit
Compare a seven-day forecast with actual results for that same seven-day period, or a monthly forecast with the month that followed. Keep the point at which the forecast was made visible, since a forecast made one day ahead is not comparable with one made several months earlier.
Use a simple absolute error and a percentage measure only when the denominator is meaningful. A very small forecast can make percentage error look extreme; include the room or revenue volume so managers understand the scale of the miss.
Break down error by date and segment
Review arrival date, channel, market segment, rate category or length of stay when the data supports those distinctions. A total may look accurate because overestimates and underestimates cancel each other, even though the hotel made poor decisions on important dates.
Cornell hotel revenue-management research examines forecasts at different levels of aggregation. The appropriate detail depends on data volume and operational decisions; adding segments without enough reliable observations can make the forecast noisier rather than more useful.
Explain the gap using evidence
Look at booking pace, cancellations, price changes, group business, room closures, events and market conditions. Separate factors the hotel could influence from external variation, while avoiding an after-the-fact explanation that has no supporting record.
Distinguish forecast error from an operational change. A renovation that removes rooms, a group cancellation or a major system outage may make the original assumptions obsolete; record when the team learned about the change and how it responded.
Check for bias and systematic misses
Calculate whether forecasts repeatedly overstate or understate demand by period, segment or lead time. A consistent direction can affect staffing, purchasing and pricing even if average error appears modest. Review both magnitude and direction.
Compare with a simple baseline, such as the same period last year adjusted for known changes, before claiming that a sophisticated model is better. Keep the baseline limitations visible and do not assume historical demand was unconstrained by room availability.
Use the review to improve a decision
Agree what the next forecast cycle will change: an input source, booking-pace rule, event assumption, room inventory adjustment or review frequency. Assign an owner and measure whether the change improves decisions over a meaningful period.
Tourism of Portugal’s municipal indicators can add local demand context, while the property’s own reservations and availability remain essential. National or municipal data should not be presented as a forecast for one hotel without a model and property-specific evidence.
Frequently asked questions
What forecast error is acceptable?
There is no universal threshold. Set a tolerance around the decision being made, forecast horizon, data volume and cost of over- or under-preparing. Review bias and operational impact alongside a numerical error metric.
Should I compare forecasts with bookings or actual occupied rooms?
Use the outcome that matches the forecast’s purpose. A demand forecast may compare with unconstrained demand if that is observable; an operating plan may focus on actual occupied rooms. State the definition and any capacity limits.