Nothing about that moment reaches a report. The stock system says the product is available, because somewhere in the network it is. The sales system shows no sale, because there was no sale to show. The customer reached a verdict about the brand, and the business recorded nothing at all.
I spend a lot of time watching how retail actually works. Not in reports. In stores.
One observation keeps coming back to me. In fashion, the purchase is personal.
When a customer picks up a pair of shoes, they are not evaluating features. They are recognising something. A fit. A feeling. That is why, when their size is missing, they do not pick an alternative. They leave. Not frustrated. Just done.
In stock is not a yes or no question
Availability is stored as a binary. The product is in stock or it is not. That is how the data model works, and for most of the supply chain it is a reasonable simplification.
It is not how the customer experiences it.
A customer does not shop the assortment. They shop one product, in one size, in the store they walked into. For them, availability has already collapsed to a single question: is my size on this shelf. Six of ten sizes present is not 60% available. For the person standing there, it is either yes or no, and four out of ten of them get no.
The gap between those two definitions is where the loss lives. The network shows healthy coverage. The shelf is failing a specific customer, in a specific location, in a way no metric is built to detect.
This is what we call SKU-level imbalance: the same product selling in one store, sitting untouched in another, and out of stock in a third, hidden by the average across the network.
A coastal resort, a capital city, and the same running shoe
The consequence is not only a missed sale. It is a misread signal, and the misread signal drives a decision.
Two locations carried the same running shoe. A capital city store and a coastal resort store.
The capital had stock. It moved slowly.
The resort sold through its sizes inside a month, and then went quiet.
Read the sales curve at the resort and it looks unambiguous. Sales rose, then stopped. Demand for this product has faded here. The standard response follows directly: delist the product from that location, clear the remainder, move on.
That decision is not careless. It is the correct reading of the data available. Sales did stop. The curve is real. What the curve cannot show is why they stopped, because a store with no stock and a store with no demand produce exactly the same line on a chart.
The YDISTRI model read it differently. The resort was not a failing store. It was an undersupplied one, with demand that had already proven itself and nothing left to sell.
So the paradox holds: a store with real demand looks like a store with no potential, precisely because it has no stock.
What changes when the stock moves
We transferred units from the capital store to the resort. They sold immediately, at full price.
That is second allocation, the method YDISTRI is built on: the corrective layer that runs after the initial distribution has filled the stores, moving stock store-to-store so the right product sits where demand has already appeared. First allocation is a forecast. Second allocation is a correction, made with the sales data the season has produced.
Without it, three things happen in sequence, and none of them appear as a single visible event. Goods reach markdown or outlet before their selling life is over. The store loses weeks of full-price selling time it will never recover. Demand goes unmet, silently, and gets recorded as absence of demand.
With it, the size reaches the location that was already asking for it. Sales recover without a promotion, a campaign, or a price cut. Discount goes back to being a last resort rather than the default response to a slow line.
This is not only a fashion problem. At CVS, stock redistributed with YDISTRI reached 83% sell-through at standard pricing over three months. The comparable stock that stayed where it was reached 1.6%.
Both figures describe the same products. The difference between them is location.
Now scale it
One shoe, in two stores, is an anecdote. That is not the scale any retailer is operating at.
Thousands of SKUs. Ten sizes each. Every store, every season. At that volume, the misread signal is not an occasional error. It is a constant, low-level tax on full-price sales, paid in markdowns that looked unavoidable and in customers who left without telling anyone why.
The stock is already bought. The demand already exists. In most cases the only thing standing between them is a few hundred kilometres and a decision nobody currently owns.
The sales come on their own, once the product is where the customer is.
So the question I would put to anyone running a store network: when a location goes quiet, how do you tell the difference between a product that stopped selling and a product that ran out of the sizes people wanted? If the answer is the sales curve, you are reading the same line for two opposite situations.
The Invisible Sale is Roland Dzogan's series on the sales and margin failures that hide before classic retail metrics notice them.
Retail as it actually works
Roland Dzogan is the founder and CEO of YDISTRI. He writes on LinkedIn about what store-level data reveals that network-level reporting cannot, drawn from work with retail chains across Europe and North America. Most of what appears here starts there first.