Inventory Shrinkage Hidden Costs Are Bigger Than Theft
General
June 1, 2026
The NRF's 2023 National Retail Security Survey pegged U.S. retail shrinkage at about $142 billion. That number gets quoted everywhere. What rarely gets quoted is the breakdown inside it. External theft drew the headlines. Process failures, miscounts, and clerical errors drew almost none. And yet, in the same survey, operators attributed roughly as much loss to operational mistakes as to organized crime.
This matters. Because most operators treat shrinkage like a security problem. It isn't. It's a data problem wearing a security costume.
Shrinkage is any gap between recorded inventory and actual inventory. Theft is one cause. Short-shipped pallets are another. A forklift driver grabbing the wrong SKU is another. A line worker pulling two units when the pick sheet said one is another. A technician scrapping a defective part without logging it is another.
The NRF groups these into categories: external theft, internal theft, process failures, and unknown causes. In the 2023 survey, external theft accounted for about 36% of shrink. Process failures and unknown causes combined accounted for more. Read that again. The stuff no one talks about is bigger than the stuff everyone talks about.
The U.S. Chamber of Commerce reported that small operators often can't even tell you which bucket their shrink falls into. They know their count was off. They don't know why. So they buy another camera.
Here's the claim you can disagree with. Most "stolen" inventory was never stolen. It was miscounted at receiving, mispicked at fulfillment, or never entered the system at all.
The pattern repeats across warehouse floors. A pallet arrives with 48 cases. The receiver counts 48. Two cases are actually 47 because one was short-shipped. Nobody notices until the cycle count six weeks later, when two full cases appear to have vanished. Security gets called. Cameras get reviewed. Nothing is there, because nothing was taken. The units never existed on the dock in the first place.
MHI's annual industry report has documented this for years. Inventory accuracy for U.S. warehouses using manual or barcode-based counting hovers in the 63% to 85% range depending on industry. That means on any given day, up to 37% of your SKU records are wrong. Not stolen. Wrong.
Theft is a real problem. But if you're spending $200,000 on loss prevention while your receiving team is eyeballing pallet counts, you've picked the expensive lock for a door that was never the one being opened.
The missing units are the cheap part. The expensive part is everything downstream.
When your system says you have 200 units and you have 178, you'll accept an order for 195. The order ships short. Your customer gets a partial. Your CSR spends 40 minutes on the phone. Your ops manager spends an hour on the floor confirming the count. Your finance team writes off the variance at quarter-end. Your auditor flags it. Your insurance premium goes up next cycle because your reported shrink rate is elevated.
Cannabis operators have it worse. Every state tracks seed to sale. METRC reconciliation failures can trigger audits, fines, and in some states license suspension. A 2% shrink variance at a grocery store is a bookkeeping line. The same variance at a Colorado dispensary is a letter from the MED.
Healthcare operators face different stakes. Controlled substances require perpetual accuracy under DEA 21 CFR 1304. A missing unit in a hospital pharmacy isn't a line item. It's a diversion report.
Industrial operators eat the cost quietly. A missing bearing on a maintenance shelf means a four-hour line stop at 3 a.m. when someone discovers the bin is empty and the record said it wasn't. The shrink cost is one bearing. The real cost is the line.
The instinct is to count more often. Monthly counts become weekly. Weekly become daily. Daily become constant. You hire another cycle counter. You add another barcode gun. Accuracy climbs a few points and then stalls.
It stalls because the counting itself is the source of error. Every manual count is a snapshot taken by a tired human with a scanner. Between snapshots, the inventory moves. Receiving, picking, damage, sampling, transfers. By the time the next snapshot is taken, the previous one is already stale. You're not measuring reality. You're measuring a past version of reality and hoping nothing moved.
The MHI data is consistent here: companies that double their cycle count frequency see roughly a 3 to 4 percentage point improvement in accuracy. Companies that instrument their bins with continuous weight-based monitoring see 8 to 12 point improvements on the same SKUs. The difference is the gap between measuring inventory and knowing inventory.
The operators who materially reduce shrinkage do two things differently.
First, they separate the problem. They stop treating shrinkage as a single number and start treating it as four numbers: receiving variance, picking variance, storage variance, and reporting lag. Each has a different fix. Receiving variance is a process and verification problem. Picking variance is usually a scan or confirmation problem. Storage variance is often a weight drift or count drift problem. Reporting lag is a system integration problem.
Second, they instrument the highest-cost bins first. Not all shrinkage is equal. A 2% variance on a $0.05 bolt is noise. A 2% variance on a $40 pharmaceutical is material. The operators who get ahead of shrinkage prioritize the items where variance costs the most, not the items where variance is most frequent.
Weight-based instrumentation helps specifically with storage variance and reporting lag. A scale under the bin reports the current count continuously. It doesn't wait for someone to scan. It doesn't accumulate lag between counts. It tells you the number is wrong the same day the number goes wrong.
If you want to talk through where your shrink is actually coming from before deciding what to fix, the team is available. Most operators find the diagnosis more useful than the tool.
Any gap between what your records say you have and what's actually on the shelf. That includes theft, but also miscounts, spoilage, damage, admin errors, and receiving mistakes.
The National Retail Federation's 2023 survey put total retail shrink at roughly $142 billion. Industrial and healthcare operators add billions more that rarely make the headlines.
A scale doesn't need a human to remember to scan anything. It reports the current weight on a continuous basis, so a missing unit shows up as a delta within seconds rather than at the next count.
No. The NRF's own data shows external theft is a large share, but process failures, paperwork errors, and unrecorded movement together account for nearly as much in most categories.
Count more often, count smaller areas, and measure discrepancies the same day you find them. Monthly full counts miss too much. Daily cycle counts on high-value SKUs catch problems while you can still fix them.
.png)
