Self-checkout moves the work instead of removing it

Most retail boards have now approved at least one self-checkout programme, and most of the cases they approved had the same shape. A number of transactions moves from staffed tills to kiosks, the till hours that served them are removed, and a saving appears in the second year. My position, after installing self-checkout in 46 of our stores, is that this shape is wrong in a predictable way. Self-checkout rarely removes work. It moves it: to the customer, to whoever remains on the floor and to the people who later investigate losses. A board should see where the work goes before it approves the lanes, not after the saving fails to arrive.

I am not arguing against self-checkout. Customers who want to pay for a bag of dog food and leave should be able to do so, and many of ours prefer it. My argument is about how the decision is costed and who is asked about it before it is made.

The hours in the business case

Our 2023 case was competently built. It took transaction data from the 46 stores with annual sales above C$18 million, assumed that a share of them would move to kiosks, and converted that share into 196,000 cashier hours a year, worth about C$4.3 million. It included the cost of the equipment, installation and maintenance. It did not include anything that happened after a customer paid.

That omission is common because it follows from the way the case starts. If you begin with transactions, you count what happens at the till. Everything else in the store is somebody else's budget. A cashier in our stores does far more than scan. She calls the yard to load a customer's truck, answers a question about which mineral supplement suits goats in late pregnancy, notices that a regular customer has not been in for a month and keeps an eye on the door. None of that appears in transaction data. Most of it still has to be done when she is no longer at the till.

Where the work went in a farm store

We installed the lanes between March 2024 and April 2025 and, against our finance team's preference, reinvested about 60% of the removed hours for eighteen months in two roles: yard loading, and advice at the feed and farm counters. No cashier was laid off. 117 moved into the new roles at C$1.40 an hour more, and 39 who wanted fewer hours took them. Then we measured where the work had gone.

Three findings changed my view. First, self-checkout handled 27% of transactions, but almost all of them were baskets under C$60. Customers buying feed, fencing or shavings still came to a staffed till or went straight to the yard, because a twenty-kilogram bag is not something you lift onto a scanning shelf. The work of serving our heaviest baskets did not shrink at all.

Second, the advice work grew. Customers who used to ask a cashier a quick question now walked to the feed counter, where the queue at weekends lengthened until we staffed it properly. Complaints about loading waits fell 44% once the yard was staffed, which told us how much loading the cashiers had been doing in the gaps between transactions.

Third, shrink rose. In the 46 stores it went from 1.31% to 1.52% of sales, about C$2 million in fiscal 2025, more than the C$1.8 million net labour saving. Our loss-prevention team had predicted something like this. I had not asked them early enough, which is the error I would most like other retailers to avoid. When they redesigned the lanes in 2026, with an attendant post facing the exit and weight checks at the bagging area, shrink fell back to 1.39%.

None of this means the programme failed. Front-end turnover in those stores fell from 71% to 52%, partly because the new yard and advice roles paid more and were more varied than a till shift. Customers with small baskets are served faster. But the honest account is that the first year cost us money, and that the shape of the original case made that outcome likely.

Count the moved work before approving the lanes

I would ask a board considering self-checkout to require three things in the case, and to read them before the saving.

The first is a work inventory: what the people at the till do apart from scanning, measured by observation in real stores over a busy week and a quiet one. In our stores that inventory would have shown loading calls, product questions and door supervision. In a pharmacy or a hardware store it will show different things. The point is to see them before deciding.

The second is a destination for each piece of that work. Some of it moves to the customer, and that is a legitimate choice if the company is willing to say so. Some of it moves to remaining staff and needs hours. Some of it, such as watching the door, becomes a security cost. A case that shows a destination for every task is a case a board can challenge.

The third is a shrink estimate prepared by loss prevention, not by the project team. The people who investigate losses know which product categories walk out of unattended lanes and which store layouts make that easy. Their estimate may be wrong, but it will be less wrong than an assumption of zero.

There are objections I take seriously. A case that reinvests hours may never realise the saving, because reinvested hours have a way of becoming permanent. That is why ours had an eighteen-month limit and quarterly reporting to the board. A retailer with a different product mix, mostly small baskets and little advice, will find that more of the work genuinely disappears, and its case should say so. And some executives will say that a work inventory slows the decision. It does, by a few weeks. The saving we promised and missed cost us more than that.

Self-checkout is a reasonable choice for many stores. My argument is only that a board approving it should know what it is buying: not fewer hours, but a different distribution of work between customers, colleagues and the people who count the losses. If the case cannot show that distribution, it is not ready to approve.