Forecasting covers: the restaurant version

A shop that misforecasts holds stock a little longer. A kitchen that misforecasts throws the mistake in the bin on Sunday night or turns bookings away on Saturday. Perishability makes restaurant forecasting less forgiving than retail's — and, fortunately, more predictable than most operators believe.

Forecast services, not weeks

A weekly covers number is nearly useless: Friday dinner and Tuesday lunch are different restaurants sharing a kitchen. The unit of forecasting is the *service* — fourteen a week — and each has a rhythm your POS already knows. Start every forecast from the same service's last six weeks and the same weeks last year, then adjust for exactly three things: bookings already on the sheet (the covers you know about — and their own pattern: how many walk-ins typically join them, how many book late), the calendar (payday weekend, school holidays, the match, the thing at the venue up the road), and the weather — which moves terraces, roasts and salads more than any other single force in hospitality.

Write the number down per service. As with the shop version, the writing-down is the method: a forecast that lives in the chef's head is renegotiated by memory every Sunday.

One forecast, three decisions

The covers number is the master key; three doors open from it. Prep — the standing over-prep habit we dissected in the food-waste post is really a missing-forecast problem: prep to the written number plus a modest buffer, not to the busiest night in memory. Rota — staff the curve of covers, not the opening hours. Ordering — the forecast times the menu mix is Tuesday's order, calculable rather than felt.

Score it by service, too

Track forecast against actual per service and the bias patterns appear fast — most kitchens systematically over-expect the quiet services (hope) and under-expect the odd spike (the calendar they didn't check). Both are fixable the week you can see them. Kauzio runs this loop from your till and booking data — service-level forecasts, weather and events folded in, accuracy scored and shown. But a whiteboard in the kitchen, fourteen numbers and honest ticks, captures most of the value. The rhythm is already in your data. Forecasting is just agreeing to read it.