Every operations manager I meet who runs an annual physical hates it. Two days of shutdown, a temp crew that's never seen your product, and a variance number in December that surprises everyone. Then a promise that next year you'll start cycle counting, and next year the physical happens again.

It's rarely a lack of effort. Plenty of warehouses count all year. They run a program that produces counts instead of accuracy, and after twelve months nobody trusts it enough to cancel the shutdown.

Counting isn't a program

Counting is someone walking out with a tablet and typing in whatever's on the shelf. If it's off, the system gets corrected and everyone moves on.

A program decides ahead of time what gets counted and how often. It sets a line above which a variance gets investigated instead of just adjusted. And it scores itself with a number you could show an auditor. Skip any one of those and you're just counting.

How often: two axes, not one

The textbook says ABC by dollar value: A items monthly, B quarterly, C once a year. It's a fine start, but value isn't the only thing that makes a part inaccurate. I also look at how often it moves. A $4 fastener picked forty times a week passes through more hands than a $900 assembly that moves twice a year, and hands are where accuracy goes to die.

High on either axis means frequent counts:

  • High value or high velocity: every 4 to 6 weeks
  • Moderate on both: quarterly
  • Low value and slow moving: once or twice a year
  • Anything with a variance history: back to the top until it behaves

Every active location gets touched at least once a year, even the dead ones, because slow movers are where phantom inventory hides for a decade.

Do the math before you commit. Multiply the SKU-locations in each bucket by how often you'll count them, then divide by working days. If that's more counts a day than one person can do in two hours, you've built a program that'll be abandoned by March. Stretch the intervals until the daily number is boring.

Count locations, not items

Tell someone to count item 44821 and they'll count the first place it lives and report back. Item 44821 also sits in an overflow rack two aisles over and in a returns cage nobody thinks about.

Count bin by bin instead: everything in C-14-3, whatever it turns out to be. That catches stock the system doesn't know is there and stock the system thinks is there that isn't. It also makes every count blind by default. The counter reports what's in front of them instead of confirming a number the screen just showed, and blind counts find things confirmation counts miss.

The threshold is the whole program

This is where most programs quietly die. A count comes back six units off and someone adjusts the system to match. The number's right now, and whatever made it wrong is still running.

Set a threshold low enough that a few items a week cross it. Anything over the line doesn't get adjusted on the spot. A different person recounts it first, because a good share of variances are just miscounts. Then someone spends ten minutes asking why.

The answers repeat. Receiving put away a case quantity as an each. Two similar part numbers got swapped at the pick face. A kit used components that never got backflushed. Someone pulled a unit for a sample and told nobody. A return went back on the shelf without being received. Each of those has a fix that prevents the next fifty variances. The adjustment prevents nothing.

Score it in a way an auditor accepts

The number most warehouses quote is net dollar variance, and it's the least useful one there is. Forty units over on one part and forty under on another nets to zero and tells you only that you have two problems.

Score location accuracy instead. A location counts as accurate when the item, the quantity and the unit of measure all match, with no partial credit. Divide accurate locations by locations counted.

The first month will be ugly. Warehouses that believe they're at 98 percent routinely open at 82 when they start scoring this way, and that gap is the reason the annual physical still exists. Track it weekly by zone and by counter. Zones tell you where the process breaks. Counters tell you who needs twenty minutes of training, which is a kinder finding than it sounds.

Earning the right to cancel

Auditors don't accept cycle counting because you say you do it. They want a written program covering scope, frequency, thresholds and how variances get resolved, plus a full year showing every location counted on schedule and an accuracy trend that's high and steady. Most firms want sustained accuracy in the high nineties before they'll drop the wall-to-wall, and some will run their own sample against your numbers first. Welcome that. If the program works, their sample proves it.

From the day you start, you're realistically twelve to eighteen months from that conversation. That's not a reason to wait. It's the reason to put the scoring in place on day one, because a year of counts you can't score is a year you'll have to run again.