15 warehouse KPIs that actually matter
Most KPI lists run to forty metrics with no priority, which is the same as having none. These fifteen cover service, inventory, productivity and cost — choose priorities from observed risk and evidence.
Beginner-friendly animated guide
Learn 15 warehouse KPIs that actually matter from the beginning
Fifteen warehouse KPIs with formulas, editable illustrative targets, and the order to fix them in — starting with inventory accuracy. Use the controls to study one section at a time. No previous warehouse knowledge is required.
Read the complete animation transcript
Before the list: the two rules
Rule one — never report a service metric without a cost metric beside it. Fill rate alone can be bought with inventory. Inventory turns alone can be bought with stockouts. Review both alongside data quality and the actual service promise.
Rule two — write the definition down before the first report. Most KPI arguments are definition arguments in disguise. Once three people are computing “on time” three different ways, the trend line is meaningless and the monthly review becomes a discussion about methodology.
Service — what the customer experiences
1. Line fill rate
lines shipped complete ÷ lines ordered. The everyday operational number. Agree a target with the service owner; this guide does not claim a verified sector benchmark. Partial shipments do not count as filled, and back orders stay in the denominator. Calculate it.
2. Order fill rate
orders shipped complete ÷ orders placed. An order counts only when every required line is complete. Its relationship with line fill depends on order sizes and where shortages occur; compare the measures using the same population.
3. On-time delivery
orders delivered by promised date ÷ orders delivered. Measure against the first date confirmed to the customer, not a revised one, and against delivery rather than dispatch.
4. OTIF
orders both on time and in full ÷ orders. Report it, but always alongside its two halves — the combined figure tells you something is wrong without telling you which half, and the two halves are fixed by different teams. Calculate it.
5. Order cycle time
Order receipt to dispatch, in hours. Track the median and the 90th percentile rather than the mean — the tail is what generates complaints, and averages hide it completely.
Inventory — what it is costing to provide that service
6. Inventory turns
COGS ÷ average inventory value. Use cost of goods, never revenue. Label the reporting period and costing basis. Compare trends only when those definitions are consistent. Calculate it.
7. Days on hand
days in the reporting period ÷ period turns. Use 365 only for a full 365-day period. State the inventory averaging method.
8. Inventory accuracy
locations with no variance ÷ locations counted. Investigate the nature and materiality of discrepancies, not only the percentage. Agree whether accuracy is measured by location, SKU, units or value. Size the count.
9. Ageing / obsolete stock
value of stock with no movement in 12 months ÷ total stock value. No movement is a review trigger, not proof that stock must be written off. Consider critical spares, seasonality, expiry and approved accounting policy.
10. Stock cover by class
Days of cover, reported separately for A, B and C items. A blended figure hides the fact that A items may be running at 12 days while the C tail sits at 400. Classify first.
Productivity — whether the operation is improving
11. Lines per man-hour
Split by function — inbound, picking, packing. A blended number moves when the mix changes rather than when performance does, which makes it useless as a trend.
12. Cost per line and cost per order
Total operating cost ÷ lines or orders handled. Normalising for volume is what makes these honest, and they will show a process or layout improvement far earlier than headcount does. Calculate it.
13. Dock-to-stock time
Vehicle arrival to stock available for picking. Include waiting time between stages, not just touch time. Stock in the receiving bay is capital you own and cannot sell, and it extends your effective supply lead time — which feeds straight back into your safety stock requirement.
Cost and space — the constraints
14. Space utilisation
occupied pallet positions ÷ available positions. Agree an operating ceiling based on layout, access, flow and safe staging. A single utilization threshold does not establish safe or efficient capacity. Size the building.
15. Warehouse cost as a percentage of sales
The number the board will ask for. Meaningful only as a trend, and only alongside service — it falls beautifully if you cut everything, right up to the point where the service collapse arrives a quarter later.
The order to fix them in
Use this as a review sequence, adjusting for immediate safety, service and capacity risks:
- Inventory accuracy first. Reconcile stock records so replenishment decisions use reliable information.
- Then fill rate and its causes. Classify the shorted SKUs before spending anything — in most warehouses a small group of items causes the majority of misses.
- Then turns and ageing stock. Once service is stable, take the cash out of the tail. Doing this before service is stable simply creates stockouts.
- Then productivity and cost per line. Process improvement on an operation that is still firefighting shortages never sticks.
- Review space alongside flow, because inventory policies and layout affect required capacity; verify constraints before committing to expansion.
How to report them
One page. Current month, previous month, same month last year, and target. A trend arrow beats a decimal place — nobody makes a decision on the difference between 86.7% and 86.9%, but everyone can act on four consecutive months of decline. Colour against target, not against the previous month, or you get a green board during a slow slide.
And put the definition of each metric on the back of the page. It ends more arguments than any amount of analysis.
Get the free KPI starter pack
Blank and invented sample copies, with editable targets and a reporting-month selector. No email required.
Original field application
Work the problem from evidence to closure
Practice situation
A KPI dashboard shows green results even though teams calculate the same measure differently. The useful response is not a generic target or a tick-box answer. Start by defining the exact event, location, item, time period and decision owner so that another reviewer can reproduce what happened.
Decision path
Write a metric contract for every KPI: purpose, formula, fields, owner, frequency, exclusions and response. Record assumptions before changing the process. Compare the proposed control with the current condition, test it on a limited scope where appropriate, and define the condition that would require work to stop or management to escalate.
Evidence pack
Source extract, calculation version, reconciliation, target basis, commentary, action and sign-off. Keep the source record with the action rather than copying only a summary into a presentation. A complete pack should show who reviewed the information, what was excluded, the date of the decision and proof that the action was later checked.
Supervisor walkthrough
- Observe the work at the location and ask the operator to explain the normal sequence.
- Compare the observed sequence with the written standard and system record.
- Separate an isolated exception from a repeated condition by checking recent history.
- Agree one immediate control and one corrective action with a named owner and realistic date.
- Return after implementation and verify the result with evidence, not verbal confirmation.
Common failure to avoid
Do not close the review with “team advised” or “monitor closely.” Those statements do not identify a changed control, accountable owner or effectiveness check. A useful closure states what changed, where it changed, when it was checked and what result would reopen the issue.
Method and verification
This lesson is an operational decision aid. Reconcile definitions to your contracts, system fields and approved local procedure. See the calculation and evidence method.
Editorial review: Amit Mehta · Original named review 2026-10-16; example updated 2026-11-01 · Operational education, not legal or engineering approval. How WarehouseOps checks and corrects content.