Kitchen forecasting basics: translate expected sales into tomorrow’s prep pars

Forecast error should be visible in both directions. Preparing 42 portions against demand of 35 is a +20% error; preparing 30 against demand of 36 is a −16.7% error and may cause a sell-out. Track the sign rather than reporting a 20% ‘accuracy gap’ that treats shortage and excess alike. Over six weeks, repeated positive errors point to inflated pars, while alternating signs may indicate that the weather or booking adjustment is too crude.

Build a baseline from services that truly resemble tomorrow

• History: select the same daypart and weekday across four to eight weeks. Use median covers and item units so one tour group or storm does not dominate the starting point.
• Bookings and walk-ins: convert confirmed reservations using the restaurant’s own no-show rate, then add expected walk-ins without counting the same guest twice.
• Weather and events: record adjustments as explicit percentages, such as −12% terrace covers for heavy rain or +15% pre-event mains, rather than hiding them inside a rounded guess.
• Constraints: cap the forecast at available seats, oven throughput, delivery slots, or staffing capacity. Demand above a physical limit cannot become sensible prep.

Work backwards from covers to the amount on the prep list

Create a forecast–actual loop the brigade will use

1. 1Select the comparison set. Before the weekly ordering cut-off, pull covers and item units from four to eight matching services. Mark closures, promotions, holidays, and large parties; exclude only with a written reason.
2. 2Layer known changes. Start with median covers, reconcile confirmed bookings using historical no-shows, then apply separate event, weather, and delivery adjustments. Write every percentage beside its source and cap at capacity.
3. 3Calculate item pars. Multiply adjusted covers by recent item mix, add an item-specific safety percentage, and subtract usable opening stock. Use smaller buffers for products that can be cooked in 6 minutes and larger ones for a 3-hour braise.
4. 4Close the learning loop. After service, enter actual units, stock-outs, substitutions, and edible leftovers. Review absolute and signed error weekly; change an adjustment rule only after at least three comparable misses point the same way.