Publish fewer availability promises, and keep all of them

Most retail websites now tell a customer whether a product is in stock at a nearby store. The feature is cheap to build once a retailer has a stock ledger and a store locator, and it tests well: shoppers like knowing before they drive. So the line appears on every product page for every store, with a confidence it has not earned. In my experience the figure underneath it is wrong far more often than anyone in head office believes.

I have seen this from three sides. As head of online product at a Toronto department store in 2012, I watched a ship-from-store pilot cancel 18 per cent of its orders because the item was not where the system said. At a software company in Montréal I spent seven years building tools for retailers who wanted to fix the same problem. And when I joined Sablemere in 2021, our own click-and-collect orders were cancelled 11.2 per cent of the time, and spring cycle counts put store record accuracy at 74 per cent. My argument here is simple, though it costs money: a retailer should show fewer availability claims, decide explicitly which ones it can keep, and treat the rest as a promise it has not yet earned.

A count is not a promise

A stock ledger records what the company believes it holds at each location. The belief drifts from the shelf for ordinary reasons. Cartons are scanned as received but not unpacked. A jacket tried on in a fitting room ends up on the wrong rail. A return is processed at the till and put in the stockroom rather than back on the floor. Theft and damage go unrecorded until the next count. None of this is negligence. It is what happens in a busy store whose ledger is updated by people doing several other jobs at once.

When the ledger was used only to reorder stock, small errors were tolerable because they averaged out across a season. Once the same figure appears on a product page, every error becomes a particular customer standing at a particular counter. The data has not changed. Its use has, and the standard it must meet has risen with it.

So the first decision is not technical. It is to accept that showing a number to a customer is a different act from holding it in a system, and to set a threshold for when that act is justified. At Sablemere the threshold is that a store's counts for the product category must have been at least 95 per cent accurate over the previous eight weeks, and at least three units must be recorded. Below that, the page offers delivery to the store in three to five days or invites the customer to call. When we introduced the rule in 2022, 41 per cent of store-and-product combinations lost the claim overnight.

Who pays when the page is wrong

The objection I heard most was that hiding availability loses sales, and it does. Conversion on the affected pages fell about 6 per cent, and we estimated the forgone orders at C$9.8 million in the first year. Our head of stores feared that customers would see empty shelves online and stop visiting. Those are real costs, borne by the company and visible in its weekly sales report.

The costs of a wrong claim are borne by other people and appear nowhere in that report. The customer who drove forty minutes on a February evening pays in time and fuel. The associate who has to search the stockroom, apologise and offer a refund pays in a difficult conversation, and in the morale that drains away after the tenth such conversation in a month. The care team pays in calls. The company eventually pays in trust, but slowly and diffusely enough that no one connects the loss to a single line on a product page.

A product leader's job, as I understand it, is to bring those hidden costs into the same room as the visible ones. We began reporting failed collections beside sales in the weekly trading review, with the number of customers affected and, where we could estimate it, the distance they had travelled. The conversation changed when store managers could see their own failures beside their own revenue. Several asked for more frequent counts before we asked them.

Earning the claim back is the second half of the rule, and the more important one. A threshold that only removes claims tells stores they have failed. A threshold that restores claims store by store and category by category, as accuracy rises, gives them something to work towards. We counted the 600 highest-selling styles weekly, fixed the receiving step where most errors began and fitted RFID tags to outerwear and footwear in 2023 and 2024. By 2025–26 record accuracy was 93 per cent across the chain and 98 per cent in the tagged categories, 88 per cent of store-and-product combinations carried the claim again, and cancellations had fallen to 2.4 per cent. Collection grew from 17 to 24 per cent of online orders, because customers had learnt that it worked.

Questions a board should ask about availability

Directors rarely see store record accuracy, and they almost never see failed collections. Both belong in the pack of any retailer that sells through stores and online. I would ask four questions.

First, what share of the availability claims shown to customers last month were wrong, measured by failed collections and by sample counts? A management team that does not know has decided by default that the figure does not matter.

Second, who owns the rule that decides when a claim is shown? If it belongs to the e-commerce team alone, it will drift towards showing more, because that team is measured on online sales. At Sablemere the rule is reviewed every quarter by stores, supply chain and product together, and none of us can loosen it alone.

Third, what is the plan to restore claims, and in which categories first? Technology such as RFID helps, but it is expensive, and its value is highest where errors cost most. Our outerwear had the highest prices, the most fitting-room movement and the most cancellations, and I regret that we did not tag it a year earlier.

Fourth, what does the customer see when the claim is withheld? We showed 'ask the store' for months without saying why, and some customers read it as 'out of stock'. A sentence of explanation would have cost nothing.

None of this is an argument against showing stock. It is an argument for treating each availability message as a promise made on behalf of a person in a store, and for showing only the promises the company has given that person the means to keep. A retailer that publishes fewer of them, and keeps every one, will find that customers trust the ones that remain.