Picture a 60,000-square-foot warehouse at 6 a.m. Forty people with clipboards, two days of counting, and a Monday morning where the system says 340 units of a fast-moving part and the shelf holds 212. That gap isn’t a computer problem. It’s a process problem wearing a computer’s costume.
The distributor I worked with had $18 million in revenue, 4,200 active SKUs, and a system that had been running for two years. Their finance team had quietly given up on inventory reports and was doing manual write-offs every Friday afternoon.
Measure one number before you change anything
Most managers want to jump straight to handheld scanners and new modules. Don’t. Start by measuring inventory record accuracy: pick 50 SKUs at random, count them by hand, compare against the system. If fewer than 48 match, you have a counting problem, not a technology problem.
That distributor scored 71%. Which means roughly one in three times a salesperson promised a delivery date, they were guessing. Customers noticed before management did.
The five places errors are actually born
- Receiving before paperwork: goods hit the shelf at 2 p.m., the receipt gets typed at 5 p.m., and someone sells three of them at 3 p.m.
- Returns parked by the door: a customer sends back 12 units, nobody posts them, and the system keeps selling stock that’s physically back in the building.
- Damage never written down: a forklift clips a pallet, the note goes on a scrap of paper, the write-off never gets entered.
- Substitutions at pick time: the picker grabs the 6-inch version because the 8-inch is buried, ships it, and the system still thinks the 8-inch is gone.
- Unit-of-measure confusion: buying in cases of 24, selling in singles, and someone picking the wrong one at 4:45 on a Friday.
Run a 15-minute count every single morning
Forget the annual wall-to-wall count. It costs you two days of shipping and buys you accuracy for about one afternoon.
- Pick 15 SKUs each morning: ten high-value or fast-moving, five completely random.
- Freeze only those 15 locations. Everything else keeps moving and shipping.
- Have a second person count independently. Two counts, no discussion, then compare.
- Log every mismatch with a reason code: receiving, picking, damage, return, unit of measure.
- Fix the process behind the number. If three mismatches in one week trace back to unposted returns, the count was just the messenger.
What 90 days actually bought them
Accuracy went from 71% to 96%. Pick errors fell from 4.1% of orders to 0.6%. Expedited freight — the money you burn when you promise a date you can’t hit — dropped by $11,000 a month.
Nobody bought new software. They changed who counts, when they count, and what happens after a mismatch. This is the unglamorous half of ERP Consulting that rarely makes a slide deck, because process discipline doesn’t come with a license fee attached.
Tomorrow morning, pull the 20 SKUs with the highest annual usage and count them by hand. Whatever you find, write down the reason. That one habit is where the recovery starts.