What is the belief, and where does it come from?
The belief is that demand forecasting is only for stores with years of sales history, and that a younger store or a new product line should reorder by feel until the data accumulates. It comes from real experience: a forecasting feature switched on with a few months of history produced numbers that were confidently wrong, the store bought to them, and the stockouts and dead stock that followed taught a lasting lesson.
The belief deserves respect because the tool was wrong to look confident. The question is whether the forecast was the problem or the hidden uncertainty was.
What is true in the belief?
What is true is that history narrows a forecast. A SKU with two years of sales reveals its seasonality, its response to promotions and its ordinary week-to-week variance, and a forecast built on that can be tight. A SKU with three weeks of sales reveals almost none of that, and an honest forecast for it is wide. It is also true that a tool which presents the wide forecast as a single number, with the same visual confidence as the tight one, will cause exactly the damage the belief remembers.
It is also true that some products have no usable history at all, a launch in a new category, and that for those the first orders are a judgment the owner makes with the forecast as one input.
What does history actually buy?
History buys a narrower interval around the forecast, not the right to forecast. A three-week SKU can be forecast; the interval is wide, and the useful output is the interval, not the midpoint. The Inventory FTE on ftes.ai labels every forecast by confidence derived from the length and stability of its history, and changes what it does with a low-confidence forecast: it proposes a smaller order, sooner, so that the store learns from a short cycle rather than committing a season's cash to a guess; it flags the SKU for the owner's knowledge, a planned campaign, a comparable product; and it upgrades the confidence as the history accumulates.
The mechanism is the label and the behaviour it triggers. A forecast that looks the same at every confidence is the myth's true cause.
How can a new SKU be forecast honestly?
A new SKU can be forecast by borrowing, honestly labelled, from SKUs the store already sells that resemble it: the same category, the same price band, the same launch pattern. The Inventory FTE proposes the comparable set, shows the owner which SKUs it borrowed from, and labels the forecast as borrowed until the new SKU's own sales take over. The owner can change the comparables, because they know the product and the data does not yet.
Borrowing is not guessing. It states its source and its confidence, and it is replaced as the SKU earns history. The dishonest version is a number with no source and no label.
What does the Inventory FTE do differently from a thin forecast that hides its width?
The Inventory FTE differs from a tool that hides its uncertainty in three visible ways. Every forecast carries a confidence label on every screen it appears on, including the purchase order proposal it drives. Low confidence changes the proposal: smaller quantity, shorter cover, an explicit request for the owner's knowledge. And forecast accuracy is reported monthly per SKU, with misses named, so a forecast that was wrong is seen as wrong and the confidence model learns from it rather than the store learning from a stockout.
None of this requires years of data. It requires stating what the data supports, which is a property of the method, not of the history.
How can a store check for itself?
A store can check the belief on any forecast by asking two questions. First, how wide is the interval around this number, and does the tool show it? A forecast that cannot answer is hiding its uncertainty. Second, what would narrow it: more weeks of sales, a comparable SKU, knowledge of a campaign? A forecast that can answer is usable now, with its width; a forecast that cannot should not drive an order at any history length.
What remains true is that a brand-new category with no comparables is an owner's judgment, and the honest forecast for it says so and proposes the smallest order that will generate the data.