Inventory Accuracy Formula
The inventory accuracy formula warehouses actually use is simple: how many counted locations matched the system, divided by how many locations you counted, shown as a percentage. Fancy finance versions that weight by value have a place later. If your floor cannot trust bin balances, start with location match rate from Stock counts.
This article defines the terms, works an example with clear numbers, and shows how to use the result beside Transfers, Scan to deduct, and multi-site on-hand without kidding yourself.
The formula
Inventory accuracy (%) = (Locations matched ÷ Locations counted) × 100
- Locations counted: distinct bins (or counted SKU-locations) you physically counted in the period.
- Locations matched: those counts where physical quantity equalled system on-hand within your tolerance (often zero tolerance for discrete cartons; sometimes ±1 for high-volume loose each).
- Locations with variance: counted − matched. These need investigation before or when you adjust.
Example policy: for cased goods like FLR-120, match means exact quantity. No “close enough”.
What not to use as your only KPI
- Company on-hand that “feels right” after netting Main DC and Van 2. Site errors cancel in totals and hide risk. Keep site splits as in track stock across multiple warehouses.
- Adjustments posted ÷ SKUs without saying how many locations you counted. Heavy adjusting can look “active” while accuracy is poor.
- One annual physical percentage with no weekly cycle sample. Accuracy is a trend, not a ceremony.
Worked calculation
Week of 5 May at Main DC. The team runs Stock counts on pick faces and two bulk locations.
| Metric | Number |
|---|---|
| Locations counted | 50 |
| Locations matched (physical = system) | 46 |
| Locations with variance | 4 |
| Inventory accuracy | (46 ÷ 50) × 100 = 92% |
The four variances:
| Location | System | Counted | Root cause after investigation |
|---|---|---|---|
| A-03-B-02 | 40 | 37 | Missed Scan to deduct on SO-9041 (3 cartons) |
| B-01-A-02 | 8 | 10 | Putaway of 2 still on dock in system elsewhere; wrong bin physically |
| Van stage cage* | 0 | 2 | Transfer TR-118 partial: 2 still in transit, staged early without receive |
| C-02-D-04 | 25 | 24 | Damaged carton removed, never adjusted |
*If Van stage cage is a counted location at Main DC, it counts in the 50. Do not also count those 2 as Van 2 on-hand until the Transfer receive posts. Mixed status is how accuracy maths and transfer SOPs collide; keep transfer states clean.
Corrected story for the week: 46 matches out of 50 locations = 92% accuracy. After fixing documents (missed scan, Transfer receive, damage note) and re-counting the four, you might hit 50/50 on a follow-up sample. That follow-up is a process check, not a licence to rewrite the original 92%.
Optional value-weighted accuracy
Once location match rate is stable, add a value view:
Value accuracy (%) = (1 − |Σ extended variance value| ÷ Σ counted extended value) × 100
Worked fragment: four variances total £180 absolute extended variance; counted locations represent £12,000 of stock. Value accuracy ≈ (1 − 180/12000) × 100 = 98.5%. You can look “accurate” on value while 92% of locations match if errors sit on cheap SKUs, or the reverse if one high-value bin is wrong. Report both when leadership only watches money.
Targets that fit an SMB warehouse
- A-class locations: aim ≥ 98% match on weekly counts.
- Whole site sample: aim ≥ 95% on the monthly mix of A/B/C.
- Go-live month: expect lower (90 to 93%) while Transfers and Scan to deduct habits settle; do not freeze the target at that level.
Publish the formula on the wall so “we are pretty accurate” stops being a feeling. Tie misses to the investigation checklist in how to investigate and resolve stock discrepancies before anyone celebrates a posted adjustment as a win.
How to gather the inputs in VORQ
- Run Stock counts by Storage site on a schedule from your cycle count plan.
- Export or tally counted locations and whether each matched.
- Exclude locations you did not count from the denominator. Accuracy is not “matched ÷ all bins in the building” unless you truly counted all bins.
- Review in-transit Transfers separately. In-transit stock is not a matched pick-face location.
- Watch incoming Purchase orders by destination in analytics so planners do not treat expected receipts as on-hand when reading availability beside the accuracy KPI.
Using the number operationally
- If accuracy drops after a process change (new van route, new receiver), audit Transfers and receiving first.
- If one aisle always fails, fix location labels and putaway rules before buying more hardware.
- If match rate is high but customer shortages continue, you may be counting the wrong locations (bulk matches while pick faces drift).
- If batch-level dispatch errors drive variances, tighten labelling with dispatch from a specific batch.
One-page summary
Counted 50 locations, matched 46, accuracy 92%. Investigate the 4. Post Stock counts only for true loss after the checklist. Next week, count A-class again and compare the percentage, not the vibe. That is the inventory accuracy formula in warehouse practice: matched over counted, every time.
When you want Stock counts, Transfers, and site-level on-hand feeding the same story, start at signup or compare pricing.
