Warehouse KPIs

Last updated October 21, 2026

A warehouse can report dozens of numbers. Only a handful tell you whether customers get the right goods, on time, at the cost you expected. Picking those few is the easy part. Defining them tightly enough that two people working from the same data get the same answer is where most of the effort goes, and where most disputes with a 3PL start.

Write the definition first

Every KPI needs its numerator, denominator, data source and exclusions written down. “Order accuracy of X%” means nothing until you know whether an order with one wrong unit out of fifty counts as a failed order, who caught the error, and whether complaints that arrive from customers weeks later are included.

Industry bodies help here because they publish definitions along with their benchmarks. The WERC DC Measures survey collects data on several dozen metrics, from on-time shipments and dock-to-stock cycle time to cube utilization and a perfect order index, and APQC’s benchmarking glossary defines many of the same things. Borrow their wording. You’ll have fewer arguments later, and you can compare yourself with outside figures if you ever want to.

The usual set, with formulas:

Outbound quality and perfect order

Picking accuracy is usually measured per order, not per unit, since one wrong item spoils the whole order as far as the customer is concerned. Decide which errors count (wrong SKU, wrong quantity, wrong lot or expiry, missing item, wrong label) and which detection points feed the number. An error caught at the pack-station check is an internal quality signal. An error the customer reports is the one that costs money. Track the two separately and you’ll see whether your checks are working.

On-time shipping should compare actual ship time with a commitment the warehouse controls, such as “orders received by 2 p.m. ship the same day.” Delivery also depends on the carrier, so on-time delivery belongs in its own measure. APQC’s definition of external perfect order performance, for instance, takes on-time from the customer’s definition of on-time delivery, which is a different question from whether the building made its cutoff.

Damage rate needs a clear line between goods that arrived damaged, goods damaged in the warehouse and goods damaged in transit. Write up damage found at receiving on the inbound paperwork. Otherwise it gets charged to the warehouse later, or to nobody.

Perfect order rolls the outbound measures into one. WERC describes a perfect order as delivered complete, on time, damage free and with correct documentation and invoicing. APQC’s glossary calculates perfect order performance by multiplying the percentage on time, percentage complete, percentage damage free and percentage with accurate documentation. That multiplied index is easy to build from KPIs you already have, and it shows how small failures compound. The alternative is a straight count: mark each order pass or fail on every criterion and divide perfect orders by total orders. You need order-level data from several systems for that, but you get a record of what happened to each order. Pick one method and put it in writing.

Receiving and inventory accuracy

APQC defines the dock-to-stock cycle as running from delivery of the goods until they are put away and recorded in the inventory management system. On the floor, choose one start event, whether gate check-in, door assignment or receipt in the WMS, and stick with it. Report a percentile or the share within target as well as the average. A few pallets left in a staging lane for days vanish inside a mean.

Inventory accuracy is most useful measured by location through cycle counting. Count a location, compare it with the system and score it correct only if every SKU, lot and quantity matches or falls within an agreed tolerance. WERC’s survey lists inventory count accuracy as “Percent by Location.” Unit-based and value-based figures have their uses in finance, but they let overages and shortages cancel each other, and a site can report near-perfect net inventory value while hundreds of locations are wrong.

Labor, space and cost

Units or lines per labor hour depend entirely on which hours go in the denominator. Settle whether indirect time (training, meetings, waiting for work), supervisors, temps and overtime are in or out. Compare like with like. A site picking full cases and one picking single units from bins will never look alike, and that gap isn’t a performance problem.

Space utilization has two common forms. Floor utilization is the share of the building’s footprint holding inventory. Cube utilization is the share of usable storage volume that’s filled, so it picks up rack height and half-empty locations. Watch for honeycombing, where locations are occupied but mostly empty; WERC tracks a honeycomb percentage alongside average and peak capacity used. Don’t chase very high utilization for its own sake, either. A nearly full building slows putaway and picking.

Cost per order or per unit is comparable only when the cost scope is fixed. Say whether it includes rent, equipment depreciation, management overhead, packaging, systems and outbound freight. When comparing 3PLs, run each provider’s rates against the same volume profile so the denominator is identical.

Putting KPIs in a 3PL contract

A service level agreement turns these definitions into obligations. Keep the list short, four to six measures with targets, and report the rest without tying money to them. Put the exact definitions in a schedule to the contract: formulas, timestamps, data source, measurement period and exclusions. Spell out what you have to supply for a target to apply, such as advance ship notices, order cutoffs, forecasts and packaging specifications. Agree on monthly reports, a regular performance review and the right to audit the underlying data. Then set the consequences: service credits or incentives, cure periods, and what repeated misses allow, up to escalation or termination rights.

Check how the KPI regime ties into the storage terms. The IWLA standard contract terms say the warehouse isn’t responsible for chargebacks of any kind and limit liability for mis-shipments, and in the US UCC § 7-204 lets a warehouse cap its liability for loss or damage by contract. If you expect retailer chargebacks caused by picking or labeling errors to be passed through to the provider, negotiate that and write it down. For how storage and handling work under public terms, see public vs contract warehousing, and use the facility checklist for site checks before signing.

Set targets from a baseline. Spend the first months only measuring, then agree on targets the operation can sustain. Watch for the usual ways numbers drift away from reality. If the provider’s WMS is the only data source and nobody audits it, agree on audit rights and sample checks. Excluding late orders the customer caused is reasonable; excluding everything that went wrong isn’t, so list exclusions by name. Cycle counts that skip high-velocity or messy locations inflate accuracy. A productivity target with no accuracy target alongside it buys speed with errors. And accuracy by unit and accuracy by order are different metrics, so don’t compare one with the other.

Published benchmarks can help in a negotiation if the definitions match. WERC’s annual DC Measures report and APQC’s open-standards benchmarking both publish performance distributions and define their metrics in detail. Before quoting a benchmark at a provider, confirm the formula, the period, the industry mix of respondents, and whether the figure is a median or a top-quintile value. A cold storage or cross-docking operation, a pallet-in, pallet-out site and an e-commerce fulfillment center will post very different numbers for the same KPI, and none of them is wrong.

If something here is wrong or out of date, please let us know.