Last month, a 3PL operations director showed me two reports. Same week, same building: Report A said pick UPH was 85, Report B said 112. He asked me: Lao Mi, which one's right? Is the system broken?
Five minutes of reading later I told him: both are right, and nothing's broken. Report A divides by pure pick time; Report B divides by clocked hours. Different denominators, different numbers. Obviously.
After a WMS go-live, the most common "system problem" is actually a people problem — nobody ever aligned on metric definitions. Everyone's UPH, accuracy, and on-time rate means something different. Let me take apart the three most-used metrics, once and for all.
UPH: change the denominator, move the number 30%
UPH (units per hour) is the most-used warehouse metric — and the messiest. I've seen at least three formulas:
Formula one: pure task time. Numerator is units picked; denominator runs from the first pick task to the last. Breaks, bathroom, waiting for waves — all excluded. Prettiest number of the three; UPH sails past 100.
Formula two: clocked hours. Denominator is shift time — eight hours means eight hours. The most "honest" and the ugliest: same people, same work, UPH drops 20–30%.
Formula three: system login time. Denominator is handheld-terminal login duration. Looser than formula one, stricter than formula two, number lands in the middle.
Three formulas, one team, three numbers: 85, 98, 112. The boss takes 112 to benchmark against the industry and feels like a champion; HR takes 85 to set piece rates and feels labor is too expensive. Talking past each other — all because nobody aligned the denominator.
My recommendation: manage internally on formula two (clocked hours), because that's the time you actually pay for. Use formula one for external benchmarking and vendor SLAs, but always label the basis. Most important: the whole company recognizes one number. Print the denominator definition next to every metric on every report — one line of small text saves endless arguments.
Accuracy: which stage you measure changes the number by an order of magnitude
"Pick accuracy 99.5%" sounds fine. But ask: accuracy at which stage?
Pick-stage accuracy: mispicks ÷ total units picked. Usually the prettiest number, because downstream audit catches some errors.
Audit-stage accuracy: orders with errors caught at audit ÷ total orders. Lower than the first, because it counts errors that survived until audit.
Customer-complaint accuracy: customer complaints of mispicks and shorts ÷ total orders. The ugliest — typically just above 99% — but the only one the customer cares about.
Three stages: 99.8%, 99.5%, 99.1%. All called "accuracy," differing only in where the error was caught. Plenty of warehouses promise clients an SLA on the first number while clients measure the third — and it ends with the lawyers.
Do it right: track all three, but use them separately. Stage one evaluates pickers, stage two evaluates the audit process, stage three goes into client SLAs. Show all three side by side in the WMS report and you can see at a glance where errors leak through — that's what data is for.
On-time rate: without written rules, the KPI means nothing
"Order on-time rate 98%" — on time against what? Wave cutoff? 24 hours after order placement? Carrier cutoff?
The worst I've seen: the warehouse defined "on time" as "shipped same day" while the client defined it as "carrier pickup scan." In between sit TMS handoff and carrier pickup delays. The warehouse believed it was at 99%; the client measured 94%. Both sides felt cheated.
An on-time definition needs four elements: start (order time or wave release?), finish (ship, handoff, or delivery?), grace (how many minutes before cutoff still counts?), exclusions (do customer-caused delays and bad addresses count?). Write all four into the SOP, pull WMS data on that logic, and only then does the KPI mean anything.
One subtlety: does the denominator include orders that never had a chance? An order placed at 10 PM against a 5 PM carrier cutoff is "late" by destiny. Include it and the metric is permanently dragged down; exclude it and the rule gets complicated. My approach: show two numbers — "achievable-order on-time rate" and "all-order on-time rate." The first evaluates operations; the second shows the whole picture.
Align definitions before building reports
Many warehouses go live on WMS and immediately start customizing reports. I say do the opposite: hold a meeting, lock down the definitions of your ten core metrics on a single page — a metric-definition handbook — then have IT build reports to the handbook. Reverse the order and every new report just creates new arguments.
The handbook doesn't need to be fancy. Four lines per metric: name, numerator, denominator, exclusions. Pin it on the meeting-room wall; every new supervisor reads it on day one. Don't underestimate that one page — in the warehouses I've worked with, the ones that aligned definitions cut cross-functional meeting time by roughly a third. Because people finally argue about "how to improve" instead of "whether your number is right."
Which formula does your warehouse use for UPH? Compare notes in the comments and see how far apart we are.
