What it finds, and how it counts.

Items fall off replenishment quietly. A store stops ordering: a de-trait, a reset, a forgotten price cut. Nobody calls. The item's sales go to zero in that store and stay there, and the loss hides inside a healthy-looking total.

The platform runs the loop. Each alert names the owner and, where there is one, the setting to fix and the ask to send. Then the comeback watcher tracks whether the store actually starts selling again.

Why a buyer believes the number. The attribution is gated and argues against itself. A comeback that predates tracking is excluded. A comeback the retailer drove on its own is credited to them. If the ask was flagged but never sent, no credit. Of $240K in comebacks, those gates hand $209K to history and to the retailer before we count a dollar.

In May, the retailer expanded one item from 210 stores to 900, overnight. On paper every new store was fully set up: traited, valid, ready. But only 220 stores were selling. Roughly 680 brand-new points of distribution had no product flowing. No angry call comes for that. The item simply posts zero-dollar weeks in stores just granted, and those weeks sit on the item's scorecard at the next review.

The platform named what actually broke. Not the trait, not the valid flag: the product was never set to flow to the store. Replenishment had never activated, so the retailer's system built no auto-order and DC stock sat. The quick fix was pushing the cases already sitting at the warehouse out to the empty stores, sized to what each one sells. The durable fix was switching replenishment on.

Selling stores went up three and a half times; sales went up by less, because a brand-new store sells less than one that has carried the item for years. The July 4th week topped $99K, but the whole account jumped by half that week, so we leave it out of the comparison.

Retail price changes get keyed into the system without the supplier in the loop. A promotion starts. A price crosses a point shoppers actually notice, and conversion quietly drops. Either way, the first signal most suppliers get is a sales number they cannot explain.

Say a question comes in by text at 9:27am: why does the review page say the promotion is working when sales are down versus last year? The platform reconciles the numbers on that page.

The promotion bought volume, not money. The price came down 15 percent and units per store rose only 8 percent, so weekly sales dollars actually fell. Margin dollars per store fell 36 percent, $25 to $16 a week. The control ran the other way: eight weeks earlier a price increase on another pack held units flat, up 2 percent per store, while margin dollars per store rose 38 percent. Same shelf, both directions, measured the same way.

Why a merchant says yes. The margin in these numbers is the retailer's own economics on the item, the same lens their merchant applies at line review. Restoring price is not a supplier favor.

A pack conversion re-keys everything. When an item moves to a new size, the retailer issues new item numbers, and the old item's demand history does not automatically follow. The retailer's system starts the successor nearly blind.

The outgoing pack sold 24 units per store per week, every week, across about 300 stores: months of proof. Its successor was keyed at 8. The system orders to the forecast, so stores were stocked for 8, sold about 9, and the low sales were on track to confirm the low forecast at the next review. Nothing looked broken. The new item was set up, stocked, and selling. Only the side by side with the predecessor showed the launch running at barely a third of its proven demand.

In early July the alert produced the ask: raise the new pack's store forecast to the old pack's proven rate, store by store. By mid-August the new pack was selling 22 units per store a week, back near the old pack's 24, and its forecast now sits at 21.

What we claim. The rate, not a dollar figure. Too few comparable stores were left without the change to price the recovery cleanly, so we do not put a number on it.