AI and purchasing forecasts: turn portions into ingredient needs

A purchasing forecast estimates demand; it is not a definitive list of tomorrow’s orders. It becomes useful when you can trace expected guests through portion counts, recipe quantities and what must actually be bought.

Reducing food waste also requires knowing what is already available and why food is discarded. The Swedish Food Agency provides guidance for food businesses.

Start with comparable sales

Choose days resembling the one you are planning: the same weekday, opening hours and service format. Note whether a dish sold out or an event brought unusual demand. Bookings provide evidence, but booked guests do not all order the same dish, and not every visit starts with a reservation.

Use a reasonable range for portions. Decide how the kitchen will respond at the upper end rather than cooking the full maximum immediately.

Let the recipe define the quantity

Specify whether the recipe uses purchased weight, trimmed weight or cooked portion weight. Do not mix them. If it uses finished weight, account for trimming and cooking yield.

Hypothetical example: you plan 80 portions, each requiring 180 grams of purchased ingredient. Basic demand is 80 × 0.18 = 14.4 kilograms. You choose a 1.6-kilogram buffer, making 16 kilograms. There are 4 kilograms of usable stock and a confirmed delivery of 2 kilograms. You therefore need to purchase another 10 kilograms.

The buffer is a planning choice in this example, not a recommended standard. Adjust it for shelf life, opportunities to replenish and the consequences of selling out.

Count what can actually be used

POS sales do not automatically reveal physical stock. Check quantity, quality and shelf life. Account for ingredients already reserved for another service or catering order.

Count confirmed deliveries once. Check delivery dates and pack sizes before rounding the purchase. A larger pack may cost less per kilogram but create more waste if the surplus cannot be used safely.

Use AI to examine assumptions

Vendion Analytics++ provides sales information. Vendion AI can help summarise recorded data and structure comparisons. Ask for the period and figures behind the answer.

Kitchen management needs to add current recipes, physical counts and delivery terms where these are missing. Those details determine purchasing; a weather forecast alone cannot calculate exact ingredient needs.

Review forecast errors after service

Compare planned portions with actual sales and record overproduction, unavailable dishes and discarded food. If the forecast was accurate but waste remained high, the issue may lie in portioning, storage or cooking.

Adjust the next plan according to the cause. The forecast becomes a better working tool even when the day differs from expectations. See Vendion live to start with your sales history.