August 2, 2026
Your 3PL's accuracy number, and the half of it you're not being shown.
Netting overages against shortages can turn a 72% count into a 98% one without a single pallet moving. Both numbers are arithmetically true.
6 min read
The short version
Inventory accuracy can be reported two ways from the same physical count, and the two produce very different numbers. Net variance subtracts overages from shortages, so items found in the wrong place cancel out items missing entirely. Absolute variance adds them together, because both are the same failure: the system did not know where the stock was. On a 10,000-unit count with 1,500 units short and 1,300 units over, net variance is 200 units and reads as 98% accurate, while absolute variance is 2,800 units and reads as 72% accurate. Both figures are arithmetically correct. Only the absolute one describes whether the warehouse can find your product. A shipper reviewing a 3PL's accuracy reporting should ask which method produces the number, whether it is measured by piece, by SKU or by location, and whether counts are reconciled against the customer's records or only against the warehouse's own.
One count, two numbers
Every physical count produces two kinds of error, and they are not opposites. A shortage is stock the system says you have and the shelf does not. An overage is stock on the shelf the system does not know about. Both mean the same thing about the warehouse: it lost track of where your product was.
How those two get combined is where the reporting decision hides.
- Net variance — shortages minus overages. Errors in opposite directions cancel each other out
- Absolute variance — shortages plus overages. Every error counts, whichever direction it went
What that looks like on real numbers
Take an account holding 10,000 units. A full count finds 1,500 units missing and 1,300 units that were not on the books.
Net variance is 200 units against 10,000 — a 2% error, reported as 98% accurate, and comfortably inside most service level agreements. Absolute variance is 2,800 units against 10,000 — a 28% error, or 72% accurate, which is not a warehouse anyone would sign with.
Nothing about the building changed between those two sentences. The stock is in exactly the same condition either way. What changed is which arithmetic went in the report.
The same trick works on the money. If the found items happen to be worth more than the lost ones, net shrinkage comes out positive — and a page that should read as a six-figure loss reads as a small gain.
Why an overage is not good news
The instinct is that finding extra stock offsets losing some. It does not, for three reasons, and they compound.
- An overage is the same broken process as a shortage — a put-away to the wrong location, a receipt posted against the wrong SKU, a pick that was never recorded. It is evidence, not compensation
- The two rarely involve the same product. Losing 1,500 units of your fastest mover and finding 1,300 of something seasonal is not a wash, it is a stockout plus a write-down
- You cannot invoice a customer for goods that turned up. Overages have no recoverable value, so netting them against real losses removes the loss from the report without removing it from your business
The questions that surface the method
None of this requires an audit. It requires asking how the number is produced, and a provider who will not answer plainly has answered anyway.
- Is the accuracy figure net or absolute? If the answer is a pause, it is net
- Measured by piece, by SKU, or by location? A building can be 99% accurate by SKU and badly wrong by piece
- Reconciled against our records, or only against yours? A count checked only against the warehouse's own system confirms the system, not the stock
- Cycle counts or an annual physical? A rolling count surfaces an error within days of it happening; an annual count tells you a year late and cannot say when it started
- What happens when a variance is found — is it investigated to a root cause and a location, or adjusted out?
How we report it
Counts are reconciled against your numbers rather than only against ours, which is the point at which a discrepancy becomes a conversation instead of an internal adjustment. Cycle counting runs continuously rather than as an annual event, so an error surfaces within days of appearing and can still be traced to the receipt or the pick that caused it.
Stock levels are visible to you through our customer portal rather than on request, and we would rather show it to you before you sign than describe it. Ask for the demo.
If you are moving because of accuracy at a current provider, say so at the quote stage. It changes how the account gets set up rather than being something we find out in month three.
Related