As of November 11, 2026, the most useful number in grocery retail is not how much stock sits in the backroom but how many sellable units face the shopper at the moment of the visit. Instacart is now selling that distinction as software: Inventory Intelligence for retailers and a Store Excellence Portal for consumer goods brands, both under Instacart Enterprise, with a daily store-level and UPC-level view of what is actually on the shelf rather than what the inventory system believes is there.
The aisle becomes a data source
The pitch rests on a gap that legacy systems never closed. Scan-based inventory records what entered the store and what passed the register, not what a shopper could pick up at 6pm on a Saturday. A case in the backroom, a product left in the wrong bay after a reset, or a promotional display covering a facing all read as in stock and behave as a stockout. Instacart frames this as the difference between a product that sold through and one that never reached the shelf, and it is the sort of distinction that changes what a field team does on Monday morning.
What makes the launch notable is not the dashboard but the definition. On-shelf availability counts only listed SKUs that are in the agreed assortment, in the planned place, with at least one sellable unit within reach. Stock in the backroom does not count, and a product in the wrong bay counts as unavailable for that visit. Written that way, the metric becomes auditable, which is precisely what most availability scorecards are not.
Six hundred thousand shoppers as sensors
Why the network matters more than the model
The data engine is the shopper base: roughly 600,000 shoppers visiting stores more than 15 times per store per day on average, producing over 10 million unique data points daily, supplemented by shelf imagery from Caper Cart smart carts and contextualised against more than 1.6 billion lifetime orders. That frequency is what turns a quarterly audit into a daily signal. It also explains the limitation: coverage follows where shoppers already go, so low-traffic formats and rural stores stay thinner than dense urban ones.
The substitution question
The portal tracks whether a shopper who met an empty shelf switched to a competitor SKU or simply to a different pack size within the same brand. That split matters because the two outcomes call for opposite responses. A pack-size switch is an assortment problem the brand can fix with the retailer; a competitor switch is a lost sale with a named winner, and it belongs in the next joint business plan rather than in a footnote.
What the portal measures
Stripped of the interface, the product is a small set of definitions that a category team can adopt whether or not it buys the software. Four of them carry most of the weight, and each one fails in a predictable way when it is reported without its companion column.
| Metric | What it counts | Where teams go wrong |
|---|---|---|
| On-shelf availability | Listed SKUs present, in place, sellable | Using inventory or POS data as the numerator |
| Lost sales estimate | Daily, store-level, UPC-level, in currency | Quoting it without a documented velocity baseline |
| Substitution tracking | Competitor SKU, pack size, or no purchase | Treating every switch as a competitor loss |
| Facing count | Facings present against planogram | Averaging across visits and hiding peak-hour gaps |
Two of these four are routinely conflated. Availability is a shelf fact; the lost-sales figure is a model output that depends on the velocity baseline behind it. Reporting the second without the first makes a gap look like a revenue event before anyone has confirmed the shelf was empty during shopping hours. The same logic applies to facings: a shelf that is full at 7am and empty at 6pm averages out to acceptable availability and terrible customer availability.
For brands, the practical consequence is a change in what gets measured during a store visit. Counting facings and recording gap type takes minutes and produces data that survives scrutiny; waiting for a syndicated panel takes weeks and arrives pre-aggregated. The chain that moved first in the United States, Dollar General, is instructive for a different reason: it now runs four delivery arrangements at once rather than picking one, which means its availability problem is multiplied across channels rather than solved by any single partner.
FAQ
Is on-shelf availability the same as in-stock rate?
No. In-stock rate comes from the retailer's inventory system and counts units the store believes it holds. On-shelf availability comes from a shelf check and counts listed SKUs a shopper can actually pick up in the planned place. A product can be in stock and unavailable at the same time.
What does a lost-sales estimate actually require?
A documented velocity baseline, a gap type, and a defined window. Without those three, the currency figure is an assumption dressed as a measurement, and it will not survive a conversation with a retailer's finance team.
Why separate competitor switches from pack-size switches?
They point to different fixes. A pack-size switch is assortment work the brand can do with the retailer; a competitor switch is a lost sale with a named winner and belongs in the joint business plan.
Does shelf data replace physical audits?
It reduces the interval between audits but does not remove the need to validate. Coverage follows shopper density, so thin formats and rural stores still need scheduled visits to confirm what the continuous signal implies.
