Blog · 22 September 2026
Comparing out-of-stocks across stores in a network: how to do it without distorting the measure?
In shortComparing stores on their out-of-stock rate raises two problems: the measure is not the same everywhere, and a figure used for ranking ends up being managed. ECR UK's industry measure put on-shelf availability at about 92% on fast sellers, and the ECR France / IRI barometer puts retailers between 5 and 8%: gaps exist, but they say nothing without a method. A network that wants useful comparisons sets the same counting method and the same hours, compares each store with itself first, only brings together comparable stores, separates what depends on the store from what depends on the warehouse, and uses the gaps to spread practices, not to publish a league table.
Why a raw comparison is worthless
The ECR France / IRI barometer gives an out-of-stock rate between 5 and 8% depending on the retailer; ECR UK’s measure found about 92% on-shelf availability on fast sellers. Inside one network, gaps between stores are of the same order. But two stores at 5% and 8% are comparable only if the figure is built the same way, which is rarely the case:
- the time of the count: a count at 9 am after filling and a count at 6 pm do not measure the same thing;
- the scope: all SKUs, or the 200 best sellers; with or without fresh;
- the definition: out of stock in the system, or absent from the shelf (stock-out or shelf out-of-stock);
- who counts: the department manager on their own aisle, or someone from outside.
The reference method is described in on-shelf out-of-stock rate: how to calculate it. A network that wants to compare starts by imposing it everywhere.
The ranking trap
As soon as an indicator is used to rank and reward, the people measuring it have an interest in improving it on paper. For a hand-counted out-of-stock rate, that is easy: count earlier, treat a half-empty crate as full, skip an aisle. The network believes it is progressing and loses the only information that mattered.
Three protections:
- a measure that does not depend on the person being judged: cross-counting between aisles, counting by the deputy, or a source outside the store;
- no bonus indexed on the rate alone;
- a regional manager who answers a poor figure with a question, not a reproach (presenting out-of-stocks to regional management).
The comparisons that mean something
1. Each store against itself
The first useful comparison is the trend: this store, this aisle, this week against the previous eight. It needs no assumption about the others and it is the hardest to manage.
2. Comparable stores with each other
An out-of-town hypermarket and a city-centre store have neither the same back room, nor the same number of deliveries, nor the same shoppers (out-of-stocks in convenience stores). Compare by groups: same format, same warehouse, same delivery frequency.
3. The share that depends on the store
Two stores served by warehouses with different service levels do not start equal. The honest comparison covers the share of out-of-stocks caused in store, measured with checking at receiving (supplier service level and out-of-stock rate).
4. The same SKUs in every store
Comparing availability on a common basket of a hundred everyday products, present everywhere, is more telling than an overall rate: you know what you are talking about, and the gap points to a precise aisle.
| Comparison | What it teaches | Risk |
|---|---|---|
| Raw ranking of all stores | almost nothing | managed figures |
| A store against its own history | whether it is improving | none |
| Stores of the same format and warehouse | what organisation changes | groups too small |
| Share of store-caused out-of-stocks | what the store controls | needs checking at receiving |
| Common basket of SKUs | where the gap is | basket to keep up to date |
What the gap is for
A gap between two comparable stores is a question: what does one do that the other does not? The answer is almost always a transferable practice: a 5 pm replenishment pass, a back room tidied by aisle, cycle counting done every day (a thirty-day plan against out-of-stocks). The network’s role is to circulate it: a visit to the store that succeeds by the one that struggles, rather than a league table shown in a meeting.
A worked example (hypothetical)
Take a fictional network of twelve supermarkets of the same format, served by the same warehouse. Declared rates run from 3% to 9%. The network imposes a 5 pm count on a common basket of a hundred SKUs, made by the deputy: the rates tighten between 6% and 10%, and the store declared at 3% turns out to be at 8%. Two stores remain clearly better; both have a late-afternoon replenishment pass. The network organises cross visits rather than a ranking. The figures are illustrative.
What shopper reporting changes
Shopper reports with ShelfAlert are a measure the store does not produce itself: the same method everywhere, continuously, with no count to manage. A network gets an honest basis for comparison between its stores, SKU by SKU and hour by hour, and each manager sees their own figures before their management does. 14-day trial, no payment card: see the plan for store managers.
Sources
- ECR France / IRI out-of-stock barometer: LSA, out-of-stock rate.
- ECR UK / IGD on-shelf availability measure: On-shelf availability: the case of a UK grocery retailer.
- Corsten, Gruen, “On Shelf Availability”, 2004: summary.