Here's a quiet way inventory tools lose you money: a product sells out, so your history shows low sales that month — and the forecast, trained on that low number, tells you to order even less next time. The stockout causes the under-order that causes the next stockout.
Your sales records only show what you fulfilled. If you had 100 units of demand but only 40 in stock, your history records 40. Those 60 lost sales are invisible — the customer bought elsewhere and left no trace. Every forecasting method, no matter how sophisticated, treats that 40 as the truth.
Feed a censored 40 into your forecast and it predicts ~40 next month. You order for 40, sell out again at 40, and the model becomes more confident the product is a 40-a-month seller. A genuine bestseller slowly gets starved down to a trickle — while you wonder why it "stopped selling."
The fix is to reconstruct true demand before forecasting: spot the months that were likely constrained by a stockout, and lift them to a sensible estimate of what you'd have sold with stock on the shelf. Done carefully, it breaks the loop — the forecast finally sees the real size of the product.
The danger is over-correcting. A genuine seasonal low or a slow product isn't a stockout, and lifting it would make you over-order. So a good correction is conservative: it only adjusts a month when it's far below both the same month in the other year and its neighbouring months, and it refuses to touch erratic products where it can't tell signal from noise.
Fopsis reconstructs true demand at a single chokepoint every method and every calculation flows through — using that conservative double-guard so real seasonal lows are never touched. And it's transparent: the data table lists exactly which months were adjusted and from what to what, so nothing is hidden.
Where stockouts understate demand, returns overstate it. Both need correcting before you forecast — see netting returns out of demand.
Fopsis corrects for stockouts automatically — and shows its work.
Start Free Trial