Your Shrink Number Is Missing the Cold Case
Retail turned shrink into a managed, reported, quarter-over-quarter profit driver. Product lost to a refrigeration failure never made the report.
Dollar General spent the last two years pulling 108 basis points out of shrink, and this quarter it lapped that improvement and kept going. Dollar Tree named shrink as a contributor to its profitability in the same week. Both companies can tell you what shrink cost them, what it costs now, and who owns the number.
Ask a restaurant or c-store operator what cold-case loss cost them last year and the answer is usually a shrug and a gesture at food cost.
Nobody stopped paying attention. The loss lands in a line item where nothing can find it, and it has been recurring there for years.
Shrink got a scoreboard. Cold-case loss did not.
Theft shrink got measured because someone decided to measure it. Once it had a number, it got an owner, a target, and a place on the earnings call. The measurement came first and the improvement followed.
Product lost to a walk-in going warm has none of that. It gets written into food cost, where it sits next to commodity movement, portion variance, prep waste, and menu mix. Food cost moves for a dozen reasons every period. A refrigeration failure is one small input into a number that is noisy by nature, so it never separates out, and nothing you cannot separate out can be managed.
The practical result is that an operator who loses eight thousand dollars of protein on a Saturday absorbs it as a bad food-cost week. The same operator who loses eight thousand dollars to theft opens an investigation.
The loss repriced this year
Cold-case loss did not get more frequent in 2026. It got more expensive, and by a lot.
Restaurant Brands International told investors that all-time high beef costs were running through its franchisees' P&Ls. BJ's Restaurants reported roughly 5% commodity inflation led by an expected 20% increase in beef. Brinker reported 4.4% commodity inflation, primarily beef.
Run that against a failure event. If a walk-in failure wrote off a certain amount of protein last year, the identical failure writes off 4% to 20% more this year. Nothing about the equipment changed. Nothing about the failure changed. The contents got more valuable, which means the cost of not catching it went up while the cost of catching it stayed flat.
That is the argument for instrumenting it now rather than next budget cycle. The economics moved in one direction only.
What catching it early actually looks like
Here is the part worth being precise about, because the category oversells it constantly.
Nobody predicts a compressor failure. What is detectable is drift. A refrigeration unit that is losing capacity does not fail at 38 degrees and then jump to 55. It holds temperature a little worse each day, recovers from door openings a little slower, and runs a little longer to get back to setpoint. Those changes are visible in continuous temperature data days or weeks before anything in the case is at risk.
So the honest version of "alert before product loss" has three parts:
Continuous temperature monitoring, not spot checks. A manager walking the back and reading a dial at 9am sees one moment. Drift is a trend, and a trend needs a series. This is the whole ballgame. Everything else is downstream of having the data at all.
A threshold that fires before the product is in danger, not when it already is. An alarm at 41 degrees on a cooler that should hold 38 tells you that you have a problem right now. An alert on a unit whose recovery time has doubled over two weeks tells you that you will have a problem, while there is still time to schedule the repair instead of taking it.
Door behavior separated from equipment behavior. A large share of temperature excursions in a busy back-of-house are a propped door, not a failing unit. If you cannot tell those apart you will either chase phantom equipment problems or learn to ignore the alerts, and the second one is worse. GlacierGrid sells an integrated door sensor at $10 per month per sensor for exactly this reason.
GlacierGrid monitors refrigeration and cold-chain temperature. The control side of the platform covers HVAC and lighting, and nothing in it forecasts a compressor failure. What it does is make drift visible early enough that the repair is a scheduled work order rather than an emergency plus a dumpster.
How to size the number before you spend anything
You do not need a platform to start. You need a number, and you can build a rough one this quarter with a spreadsheet.
Pull every emergency refrigeration service call from the last twelve months across the fleet. For each one, get two things from the store: what it cost to fix, and roughly what got thrown out. Managers remember the throw-outs. Add a line for any health-department or quality hold that traced back to a temperature event.
Add it up and divide by location count. That is your annual cold-case loss per site, and it is almost always larger than the operator guessed, because it has never been added up before.
Then compare it to what the same loss would have cost at current protein prices rather than last year's. That second number is the one to take into a budget conversation, because it is the one that is still moving.
What to do next
Count it first. An unmeasured loss cannot be assigned to anyone, and a loss with no owner does not improve, which is the entire lesson of what retail did with theft shrink over the last decade.
If the number comes back big enough to act on, the next step is continuous temperature data on the cases that hold the most expensive inventory. Not every unit in the fleet on day one. The walk-ins holding protein, at the sites with the worst service-call history.
GlacierGrid runs a free 90-day pilot. Across deployments, operators see roughly 10% energy savings and 15% fewer service calls, and the refrigeration monitoring is what turns a cold-case failure from a Saturday emergency into a Tuesday work order.
Start by counting what it costs you today.
