The system says eight. The shelf has none.

Every retailer has seen it. The system shows eight units on hand. The shelf is empty. Because the system believes there is stock, it doesn't reorder, and the shelf stays empty until someone happens to count it.

That can be days. For a slow-moving line in a busy store, it can be weeks.

Where the missing stock goes

Usually nowhere dramatic. A damaged unit thrown away without a write-off. A multipack scanned as a single. A receiving error at the back door. Theft. Stock that is physically in the building but sitting behind a pallet in the backroom where nobody will find it before it expires.

Each one is small. Together they add up to a store where the stock record and the shelf drift apart a little every day, and the ordering system trusts the record completely.

Replenishment only reacts to the stock it can see. Stock that exists only in the system switches reordering off.

Why this matters more than it looks

IHL Group's 2026 study of inventory distortion names phantom inventory as one of the primary drivers of the $1.7 trillion a year retailers lose to out-of-stocks and overstocks. ECR's work on retail loss puts the sales uplift from simply correcting inventory-record accuracy at 4 to 8 per cent, even in businesses that already run well.

It also poisons everything downstream. A line with phantom stock shows zero sales, so its forecast falls. When someone finally counts and corrects it, the system reorders against a forecast that has been learning from an empty shelf.

Finding it without counting every shelf

You don't need to count everything. The signal is already in your data: a line with positive stock on hand, a forecast that says it should be selling, and no sales at all for several days in a row. One day of zero sales means nothing. Five days on a line that normally sells every day means something is almost certainly wrong.

A simple testFor every store and product, flag where system stock is above zero, expected daily sales are above one unit, and actual sales have been zero for at least four consecutive days. Send that list to the stores as a count request, not a report. The first time most retailers run this, the list is longer than they expected and the hit rate is higher.

Then treat the stock as zero for ordering until someone confirms it. It is better to occasionally order a case you didn't need than to leave a shelf empty for three weeks because a number said otherwise.