Forecasting and prediction
Forecasting occupancy for a New Zealand accommodation business
Use booking pace, cancellations and MBIE guest nights to forecast occupancy. See how a New Zealand operator can test a forecast against housekeeping needs.

A motel owner needs to offer housekeeping shifts before all the rooms are booked. The reservations screen shows rooms sold so far. Last summer's accounts show how much the motel earned. Neither answers the immediate question of how many rooms will need a changeover on Sunday morning. A forecast should name the date, unit and decision it supports.
What can you tell from the bookings already on hand?
Keep a dated snapshot of future bookings each time the roster is prepared. Count confirmed rooms by arrival and departure date, and record cancellations when they occur. A final booking export shows who eventually stayed. It cannot reliably reconstruct what the duty manager knew when the roster closed unless the system preserves that history.
Suppose an invented property has 18 rooms, with two out of service, and ten rooms already booked for Saturday. If three more rooms book and one cancels, twelve of the sixteen available rooms would be occupied. Housekeeping still needs the departure list. A guest staying another night and a room being turned for the next arrival generate different work.
Track the date on which a room goes out of service and the date the desk learned about it. Treat a maintenance closure as a change in available supply. If it is simply recorded as an unsold room, the system may interpret a building repair as a drop in guest demand next time those dates come around.
How does regional tourism data help?
MBIE’s Accommodation Data Programme reports guest nights and separates domestic and international visitors. A family of four staying together creates four guest nights in one occupied room for that night. Use regional figures to understand broad seasonal movement and pressure across the area. An operator’s booking history supplies the direct evidence for their own property and guest mix.
Treat events carefully. A sports tournament may bring group bookings that appear early and cancel together. A long weekend can attract short stays booked late. Ask the supplier to show the booking curve for each type of period and record which events the manager actually knew about when preparing the forecast. A calendar updated after the fact gives an unfair advantage.
Which result belongs on the roster screen?
Show likely departures, stayovers and uncertainty by day, with an explanation if the estimate changes sharply. Set a decision point with housekeeping. For instance, the manager might hold an optional shift open when the plausible range spans a level the roster cannot cover. Use the operator’s actual staffing constraints to decide where that point sits.
Trial the forecast alongside the existing roster process through a range of ordinary and busy weeks. Compare changeovers understaffed, paid hours unused and time spent adjusting the output. Save manager overrides with a reason, such as a group moving its arrival. At handover, the owner should be able to read a forecast, identify the assumptions and run the next roster when the external tourism data is delayed.