Six months ago, a metal fabricator with 140 people on the floor had the same crisis every Tuesday. The Monday night shift would start a job, discover the brackets weren’t there, and stop the line. Forty people stood around waiting for someone to drive 90 minutes to a supplier.
Their planning screen said 1,200 brackets in stock. The shelf held 304. Nobody had stolen anything. The numbers had simply drifted, one small mistake at a time, for about four years. That gap was costing them roughly $18,000 a week in idle labor and expedited freight.
The number on the screen is not the number on the shelf
Most operations managers I meet already know their inventory accuracy is bad. What they don’t know is how bad, because nobody counts. They trust the system, the system trusts the last person who typed into it, and that person was in a hurry.
We ran a two-day cycle count on 60 of their highest-moving parts. Accuracy came back at 71%. For a shop their size, 97% or better on those items is the bar. The 26-point gap was the entire production problem — not scheduling, not the shop floor, not machine capacity.
- Pick your 50 highest-value or highest-volume parts. Not all 4,000 — just 50.
- Count them by hand, twice, on two different days, by two different people.
- Compare both counts to the system. Write the variance in dollars, not units.
- Anything off by more than 3% gets a root-cause note: wrong unit of measure, unreceived delivery, scrap never reported.
What the first 90 days actually looked like
They didn’t rip out their system. They didn’t spend much money either. They fixed the four habits that were feeding bad data into it.
- Receiving moved to the dock. Nobody walked a delivery into the warehouse and typed it in later. It got scanned at the door, against the purchase order, before the truck pulled away.
- Scrap got a two-minute form. Operators had been tossing bad parts in a bin and moving on. Now every scrapped part hits a tablet, and the bin gets emptied at shift end.
- Substitutes needed approval. When a buyer swapped a similar part to keep a job moving, the swap got flagged the same day instead of never.
- Fridays became count day for the top 50. Thirty minutes, four people, a rotating list.
By day 90, accuracy on those 50 parts hit 98.6%. Tuesday stoppages went from three a month to zero for eleven straight weeks. That’s the kind of result ERP Consulting is actually paid for — not the software, the habits wrapped around it.
The part that usually gets skipped
Pick the 50 parts this week. Count them Friday afternoon. You’ll have your answer by 4pm, and it will probably be uncomfortable.
The uncomfortable number is useful. It gives you a dollar figure to put in front of ownership, and it turns “our data is bad” from a complaint into a budget line. Do that before you call an ERP Consulting firm or sign anything.