Demand forecasting is not a crystal ball; it is a planning buffer. It turns “we think Friday will be busy” into practical prep quantities and staffing levels. Booked pre-orders are the strongest signal. Good forecasts provide a range rather than a single number, respect the kitchen’s real limits, and let managers override them for one-off events no model can see.
A demand forecast is not a crystal ball; it is a planning buffer. It turns a vague “we think Friday will be busy” into prep quantities, thawing plans, and staffing levels that hold up in real life. The point is not to predict the future perfectly. It is to avoid being surprised by a Friday you should have seen coming.
The strongest signal in any forecast is demand already booked: the pre-orders already on your books. A model that ignores tomorrow’s paid orders is only guessing — while your own system already knows half the answer.
Move from one guess to a range
The real step forward is replacing one number with a range your managers can genuinely trust. Instead of “we will do €5,000 on Friday,” a good forecast says, “with high confidence, at least €4,000; on a good day, up to €7,000.” Staff for the likely case and prepare the kitchen for the good case. That removes the scramble without buying stock for a fantasy scenario.
Tie it to the kitchen’s limits
A forecast is only useful if it knows the physical limits of your restaurant: prep times, pickup and delivery slots, and the capacity of each station. A number the kitchen cannot possibly fulfil is not a plan; it is a wish list. The forecast should therefore use the same settings that control ordering. Then it tells you not only what is coming, but also what the kitchen can actually handle.
People keep the final say
A model can process history, channels, weather, and school holidays brilliantly — but it cannot see the roadworks on the corner, the concert next door, or a public holiday with its own local rhythm. A useful forecast therefore lets the manager step in: adjust for one-off events, with the system recording the change so you can later see what was closer to reality. That feedback is how the forecast learns your location’s rhythm instead of claiming to know it from the outside.
The 7 most common mistakes
- Using only last year’s number for the same day.
- Ignoring booked pre-orders — the strongest signal.
- Using one number instead of a range with confidence.
- Failing to tie the forecast to the kitchen’s limits.
- Buying stock and staffing for the best-case scenario.
- Allowing no manual override for one-off events.
- Having no feedback loop — so the forecast never learns.
How to build a forecast you can use
Frequently asked questions
Isn't demand forecasting just guessing with a better name?+
Why are pre-orders the most important signal?+
Why use a range instead of one exact number?+
Does the forecast replace a manager’s judgement?+
Calm instead of surprises
A good forecast does not make the week predictable like clockwork — but it takes away the nasty surprise. History, events, and above all booked pre-orders create a range the team can trust, tied to the kitchen’s real limits and supported by human judgement. “We’ll see” becomes a plan, and volatility becomes a calmer operation.
