Overstock a perishable and it rots. Under-stock it and the customer walks. In food and beverage the forecast is judged on both ends at once, and the demand itself is genuinely hard to call.
Sources: Gartner, via ISM (2024); ReFED, 2025 US Food Waste Report (2023 data).
Most sectors pay for a forecast error once. Food pays twice. Order too much of a short-dated line and it is written off at the back door; order too little and the shelf goes empty, which in grocery is worse than a lost sale. A shopper who cannot find your yogurt grabs a competitor's, and there is a fair chance they do not come back for yours next week. So the planner is squeezed between spoilage on one side and a lost customer on the other, with very little room in between.
This is not a case of lazy planning. Gartner's benchmarking, reported through ISM, puts the median demand-forecast error in food and beverage at roughly 25 percent. A quarter off, on the median item. That is what happens when you stack weekly promotions, sharp seasonality, short shelf life and a catalogue that keeps sprouting new flavours and formats on top of each other. A moving average never had a chance against that.
And the waste is not marginal. ReFED's latest figures put surplus food from US manufacturing at about $41.9 billion a year, part of a problem that costs US food producers and businesses on the order of $108 billion in lost value. A meaningful slice of that is simply demand that was planned wrong.
Promotions are the clearest example. A feature or a price cut creates a short, violent spike in demand, and a moving average, by design, only notices it after it is over. Build for last month and you stock out mid-promo on the very item you are advertising. Build on a gut-feel uplift and you are often left holding the excess once the promo ends and the date is already ticking. Promotions should be planned as the demand drivers they are, not discovered after the fact.
In a long-life category, a forecast that runs 10 percent high just means a bit more stock sitting around. In fresh or short-dated food, that same 10 percent is a markdown and then a bin. The clock is always running, which is why the buffer has to be sized tight and why catching a slow mover early, while it can still be moved or discounted, matters far more here than almost anywhere else.
The teams that get this right do a few unglamorous things well. They forecast promotional and seasonal demand reproducibly, modelling the spikes instead of smoothing them flat. They size safety stock per item on that item's real variability, so the fast movers are protected and the slow ones stop quietly becoming waste. And they watch coverage on short-dated lines closely enough to act while the stock still has value. It is the same deterministic, explainable discipline DemandIntel brings to food distributors and manufacturers, running on your own data, on top of the ERP you already have.
Weekly promotions, strong seasonality, short shelf life and constant new-SKU launches all pull on the same forecast at once. That combination is why median forecast error in the sector sits around 25 percent, far higher than in steadier categories.
Forecast promotions and seasonality explicitly rather than averaging them away, size safety stock per item on real demand variability, and watch coverage on short-dated lines so slow movers are caught while they can still be sold or discounted.
Yes. Promotions are modelled as the demand drivers they are, so the plan anticipates the spike and its aftermath instead of being surprised by both. That is what keeps you from stocking out mid-promo and drowning in excess afterward.
It does. Short-dated and shelf-stable lines are planned with the shelf-life and service logic each one needs, in the same system, so one team can manage both without juggling spreadsheets.
In food, an empty shelf is not a deferred sale. It is a customer learning they can live without you.
Picture your next S&OP meeting: one number on screen that nobody argues with. See it free, on your own data, in 45 minutes; and if it does not hold up, we part as friends.