“How many of these do we actually have?” The warehouse manager looked at the screen, then at the shelf, then back at the screen. The system said 240 units. The shelf held 61.
That was a Tuesday in March. The company is a building-products distributor doing about $85 million a year out of two warehouses. Nobody thought they had an inventory problem — until that one SKU turned out to be one of about 1,900 that were wrong.
The 18% nobody budgeted for
We ran a blind count of 400 SKUs. Blind means the counter gets a blank sheet and a shelf, with no system number to copy down.
The result: 18% of counted SKUs were off by more than 5%. One in five. Finance had been closing the books on those numbers for years without knowing it.
The cost wasn’t abstract. Roughly $310,000 of stock sat in bins the system called empty, so nobody sold it. At the same time the system showed 900 units of a fast mover that didn’t exist, so sales kept promising delivery dates on air.
Stop counting everything, count the right things
A wall-to-wall annual count feels responsible and fixes almost nothing. You shut the warehouse for a weekend, find a variance, adjust the number, and three months later it’s wrong again — because nobody asked why it drifted.
Here’s what actually worked:
- Split all 4,800 SKUs into A/B/C by annual dollar movement. The 380 A items got counted monthly, B quarterly, C once a year.
- Count 15 to 25 SKUs every morning before the first truck leaves. One person, about 40 minutes, no shutdown.
- Log the reason, not just the variance. Pick errors, receiving short-ships, unrecorded returns and wrong units of measure accounted for nine out of ten errors we found.
- Recount any A item that’s off by more than 2% within 48 hours, and read the transaction history before touching the number.
The awkward part: it was never the software
By week six the pattern was obvious. Two-thirds of the errors started at receiving, on a dock where product arrived faster than anyone could key it in. People were typing quantities from memory at 5 p.m. because the line was already backing up.
The fix was a scanner at the dock and a hard rule that nothing gets put away before it’s received. It cost less than the value of one miscounted pallet, and weekly variance dropped by roughly 60%.
This is where ERP Consulting engagements either earn their keep or waste your money. Software holds the record; it doesn’t create the discipline. Buy a new system and change nothing about how the dock works, and your numbers stay wrong — just faster.
Where they landed at day 90
Accuracy on A items went from 82% to 97.4%. Write-offs on dead stock fell by $190,000 over the next two quarters, and the buyer stopped padding safety stock on items that were never actually short.
One detail is easy to skip and matters more than it sounds. They now write the accuracy number on a whiteboard in the warehouse every Monday. Not to shame anyone — as a score. People got better partly because they could finally see the number move.
Tomorrow morning, pull 20 SKUs that sell every week and count them blind. If more than two are off by 5% or more, you now know exactly where next quarter goes.