Out-of-stocks: measuring the sales you never saw

The most expensive number in retail is one that never gets recorded: the sale that didn't happen because the shelf was empty. An out-of-stock, or OOS, is a moment when an item a store normally carries isn't available to buy. Measuring it is awkward by definition, since you're measuring an absence, but scan data makes the absence visible through the silence it leaves in an item's sales history.

That's the core method: you can't observe the missed sale, but you can observe when a reliable seller inexplicably stops selling.

How does scan data detect an out-of-stock?

By baseline and gap. Every item in every store has a sales rhythm: so many units a day or a week, with normal variation. When an item that scans steadily goes silent, and stays silent past what its normal variation allows, the most likely explanation is that there was nothing on the shelf to scan. A purely hypothetical example: an item that has sold several units a day for months, then records zero for four straight days, almost certainly wasn't suddenly unloved. It was unavailable.

The inference gets stronger with the item's speed. Fast movers have tight rhythms, so a gap stands out within a day or two. Slow movers are genuinely harder, since a quiet week may just be a quiet week. Good OOS measurement is honest about that asymmetry and reports confidence accordingly.

What is phantom inventory?

The reason scan-based detection matters even in stores with inventory systems. Phantom inventory is stock the system believes exists but the shopper can't buy: units lost to theft or damage, misplaced in a back room, or miscounted at receiving. The inventory record says the item is in stock, so no reorder triggers, while the shelf sits empty for days.

Scan silence cuts through this, because it reflects the shelf as shoppers experience it. If the system says twelve units on hand and the register hasn't scanned one in a week, the twelve units are a rumor. Sales-side detection and inventory records work best checking each other.

How do you estimate the sales you never saw?

Multiply the baseline by the gap, and present it as a range. As a hypothetical illustration: an item averaging 5 units a day that went dark for 4 days points to roughly 20 units of missed sales, give or take the item's normal variation. Some of that loss is softened when shoppers substitute another size or brand, and some is compounded when a disappointed shopper takes the whole trip elsewhere. The honest estimate acknowledges both without pretending to measure them precisely.

Aggregated across stores and weeks, these estimates turn OOS from an anecdote into a managed number: which items go dark most often, where, and at what cost.

Why do averages hide the problem?

Because a monthly total looks fine after the gap is over. A store that missed four strong days can still post a respectable month, and the average smooths the wound into invisibility. OOS lives at daily, store-level resolution, which is exactly the resolution scan data preserves and summary reports discard.

Small-format retail raises the stakes. An independent store carries one facing of an item, not a case stack, so a single missed delivery can mean days of silence. For brands, watching gap patterns across the independent channel, at the store-level resolution networks like the one behind NRS Insights collect, is the difference between knowing OOS exists and knowing what it costs.