Count the shrink before you lock the shelf

Shrink became a board subject across American retail in the early 2020s, and boards were right to ask about it. My argument concerns the order of the response. In a small-box discount store, locking merchandise and hiring guards should come last, after the company has broken its shrink figure into separate causes and fixed the ones it controls. Safety should be decided separately, as a question of how many people are in the store, and should never be justified or rejected by its effect on shrink.

I began my career in 1994 as a store detective at Tollerson's Discount Centers in Joplin, Missouri, and spent most of my first twelve years in loss prevention. The lesson I learnt earliest, and have had to relearn as a chief executive, is that shrink is a ledger number. It is the difference between the inventory a company believes it holds and the inventory it counts. Theft is part of that difference. In my experience it is rarely all of it.

Four causes behind one percentage

At Marrick Variety Stores shrink rose from 2.2% of sales in fiscal 2020 to 3.1% in fiscal 2022. The rise was real and expensive: on our sales, each tenth of a point is close to US$2 million a year. The early explanation, from our own field leaders as well as from vendors of security equipment, was organised theft. Stonebeck's operating partner asked for a plan to lock high-loss categories and place guards in our worst 120 stores, at a cost of about US$6 million a year.

Before approving it, I asked our asset protection and finance teams to decompose a year of shrink in those 120 stores, using inventory audits, receiving records, markdown logs and case files. The answer did not dismiss theft, but it changed the shape of the problem. About 38% of the loss came from process: deliveries received without counting, vendor credits never claimed, damaged goods thrown away unrecorded and price changes entered late. Roughly 9% came from employees, mostly at the register. External theft, opportunistic and organised, accounted for the rest, and it was heavily concentrated: forty stores carried more than half of it.

That breakdown changed what we did first. We fixed receiving for the suppliers who deliver directly to stores, retrained store managers on recording damage and markdowns, and audited refund and void patterns at the register. Those measures cost little and removed about four-tenths of a point of shrink within a year. We put locked cases and other physical measures only in the forty stores where external theft was concentrated, and only in the categories where it was concentrated there.

What a locked case costs a store with two people in it

Locked merchandise has costs that are easy to understate in a large store and impossible to ignore in a small one. A typical Marrick store has 8,200 square feet and, for most of the day, two or three people working. When a customer needs a locked pack of razor blades or a tin of infant formula, someone must stop what they are doing, find a key and walk to the case. If one of the two is at the register, the store has nobody on the floor. In our trial stores, sales in locked categories fell by between 11% and 19%, and customers told managers they had driven to another town rather than wait.

Some of those lost sales were sales we would otherwise have lost to theft. Not all of them were. A board approving locked cases should see the expected fall in sales next to the expected fall in loss, and should know who in the store will carry the key. In a store of our size the honest answer is usually the cashier, who is also the person a frustrated customer confronts.

Store safety belongs in the staffing decision

The more serious problem at Marrick in 2022 was not shrink. It was that 61% of our stores closed at 10pm with one employee on duty for the final two hours, and we recorded forty-seven robberies that year. Treating that as a loss-prevention problem would have meant asking whether a second person paid for themselves in reduced theft. Sometimes they would; often they would not. It was the wrong question to ask.

From April 2023 we required at least two people in every store from opening until close, and moved closing time from 10pm to 9pm in 640 stores to help fund the change. The net labour cost is about US$11.8 million a year. We lost some evening sales, and in towns where a factory shift ended at 10pm we lost more than I expected; we restored the later hour in twenty-three stores with three people on the close. Robberies fell to fourteen in fiscal 2025, and assistant manager turnover fell from 78% to 51%. Shrink fell too, but I would have argued for the change without that.

Separating the two decisions matters because each distorts the other when they are combined. If safety is justified by shrink, a store with low shrink and real danger will not qualify for a second person. If shrink is used to justify spending, the loudest explanation, usually organised theft, wins the budget before anyone has counted the receiving errors. And a manager told that a colleague's presence depends on loss figures learns something about how the company values them.

There are limits to this approach. Some stores do face sustained organised theft that process cannot fix, and their staff deserve protection that may include guards and locked cases. Decomposition takes time and data that a company in crisis may not have; ours occupied six people for about ten weeks, and we had to rebuild a year of receiving records by hand. Our breakdown came from a rural network, and a dense urban chain may find external theft a larger share of its loss. I would still ask any retail board to require the decomposition before the spending, to see lost sales beside reduced loss, and to decide how many people should be in a store at closing time without reference to shrink at all.