There's a quiet problem sitting inside almost every fashion merchandising team's planning process — and most planners don't even know it's there.
It doesn't announce itself. It doesn't show up as an error. It looks, on the surface, like perfectly normal data. And yet, season after season, it causes brands to underorder their best products, overorder their worst ones, and wonder why the forecast always seems off.
The problem is called a corrupted demand signal. And the root cause is something deceptively simple: your system is planning from what sold, not from what customers actually wanted.
The Stockout Problem Nobody Talks About
Here's a scenario that plays out in every apparel business.
A linen co-ord set launches in April. It sells out completely by the second week of May — six weeks before the season ends. For those six weeks, every customer who walked in looking for it left empty-handed. No sale recorded. No data captured. Just a gap on the shelf and a lost opportunity that the system files away as zero.
Come October, when the planning team sits down to plan next year's summer season, they look at last year's numbers. The co-ord set shows four weeks of strong sales, then nothing. If they're using a tool that doesn't account for that stockout — which is most tools — the system reads the "nothing" literally. It sees a product that stopped selling in May and concludes demand was moderate. Maybe it recommends buying 20% more than last year. Maybe 30%.
What it should recommend is buying significantly more. Because the product didn't stop selling in May. The brand stopped having product to sell.
The historical record is a lie — not a malicious one, but a structural one. And if your forecasting system doesn't know how to read between the lines of that lie, every plan you build on top of it inherits the distortion.
Why This Matters More Than Most Brands Realize
The consequences of planning from a corrupted demand signal compound in both directions.
Your stars get underestimated. The products that sold out early were, almost by definition, your best performers. They had the highest real demand. But because their sell-through looks truncated in the data, the system treats them as moderate performers. Next season, you buy less than you should. You repeat the stockout. You repeat the miss. Your dogs get overestimated. Meanwhile, the products that sat on shelves for four months have pristine-looking data — no stockout gaps, consistent (low) weekly sales all the way through. The system reads this as steady, reliable demand. You buy a full range again next season. They sit again.Over time, this creates a systematic bias toward mediocrity: your buying process trends toward the products that sold steadily because they're always in stock, not the products that sold best because they're always running out. You end up with more of what nobody loves and less of what everyone wants.
What It Actually Takes to Fix It
Correcting for lost sales isn't complicated in concept, but it requires a system that's actively looking for the problem. Most platforms aren't.
The fix involves a few things working together:
Identifying the stockout windows. The system needs to flag every period where a SKU's inventory hit zero — not assume that zero sales means zero demand. This sounds obvious, but it requires your POS, inventory, and planning data to be connected and talking to each other in real time. Estimating what demand actually was. Once a stockout window is identified, the system needs to make a statistically reasonable estimate of what sales would have been if stock had been available. This typically uses the product's pre-stockout velocity, adjusted for seasonal curves and comparable store performance. Replacing the corrupted period in the historical record. Instead of feeding the plan a six-week block of zeros, the system substitutes a reconstructed demand estimate. The historical record now reflects an approximation of true demand rather than supply-constrained actuals. Flagging the adjustment transparently. Planners need to know when they're looking at corrected data versus raw actuals. Hiding the correction doesn't help anyone — seeing it builds trust in the forecast and lets experienced planners validate the estimate against their own knowledge of what happened.The result is a demand signal that reflects what the customer was trying to buy, not just what the business managed to fulfill.
The Hidden Cost of Ignoring It
Teams that don't correct for lost sales often compensate with instinct. Senior planners know which products "felt" bigger than the data shows. They add buffer quantities based on gut feel. They argue with the system output and manually adjust upward on the things they believe in.
This isn't irrational — it's a reasonable response to a system they've learned not to fully trust. But it's inefficient, it doesn't scale, and it walks out the door with every experienced planner who leaves.
The goal of a good forecasting system isn't to replace planner judgment. It's to give planners a demand signal accurate enough that their judgment can be applied to genuine strategic decisions — which new styles to back, which categories to stretch into, where to take calculated risk — rather than spent correcting for structural data errors.
When the system tells the truth about demand, planners can spend their energy thinking forward instead of wrestling with the past.
Planning From a Clean Signal
The shift from corrupted to corrected demand data changes buying in concrete ways.
Reorder decisions become more defensible. When a product is flagged as a strong performer, the team can see whether that performance is real or inflated — and when a previously understated product gets its corrected history, the case for backing it becomes clear and quantifiable.
Range decisions get sharper. If you can see which product attributes genuinely outperformed (rather than which ones happened to stay in stock longest), range architecture for the next season reflects real customer preference rather than supply luck.
And crucially, the planning process builds institutional knowledge instead of eroding it. Each season's corrected data makes the next season's forecast more accurate. The system gets better over time, which means the team gets better over time — even as individual planners come and go.
Most brands are one or two seasons away from having a demand signal they can actually trust. The first step is recognizing that what your current system calls history isn't the full story.
The best merchandising teams don't just plan from data — they plan from accurate data. There's a difference, and it shows up every February.