Here's the test I give every warehouse owner: if your WMS has been live for six months and you still have to chase your ops manager every morning asking how many orders went out yesterday, you paid for a system and got a very expensive filing cabinet.
Back in 2021 I walked into a 3PL warehouse in Ontario, California — 120,000 square feet, over 40,000 SKUs, pushing 20,000 orders a day in peak season. Three years on their WMS, full reporting module purchased. The owner asked me, "Lao Mi, tell me straight — was this system worth it?" I asked what number he looked at every day. Just one: yesterday's shipped orders. I told him at least half his software money was wasted.
What a boss needs isn't "how many shipped." It's "is anything about to go wrong today." I gave him six numbers, had IT put them on a single page, auto-generated at 6 a.m. and in his inbox by 8. Three months later he told me it was the highest-ROI tweak he'd ever made to the system — and it cost zero in new software.
Why 8 a.m.? Because after 8, the phone starts ringing and the floor starts queuing up problems. A dashboard's value is the glance before the shift starts — a morning with the numbers and a morning without are two completely different ways to run a day. I've seen dashboards built as real-time video walls in the conference room, all red and green and impressive. Three months later, nobody had looked at them twice. Owners don't need a video wall. They need five certain minutes every morning.
The three numbers that tell you about today
Ship-on-time rate. Not total orders shipped — the share of orders promised for today that actually left before your cutoff. If your same-day cutoff is 3 p.m., measure what percentage was out the door by 3. My line is 95%. Drop below it and somebody needs to be watching waves and labor for the rest of the day. It all lives in the outbound order table: apply the cutoff, compute the rate, five minutes of work.
Pick UPH against your own baseline. Units per hour per picker. Don't benchmark against industry averages — compare against your own building's history. Three straight days running 15% below baseline almost always means slotting has drifted or waves are badly grouped. The trap here is the pretty average: with ten pickers and three coasting, the average UPH can still look fine. So the dashboard shows the average plus the bottom three individuals. Naming names isn't about punishment — it's how you tell whether the problem is people or process.
Overtime and temp share. Overtime hours plus temp-agency hours as a share of total labor hours for the week. Cross 20% and your wave planning or headcount model is broken — labor cost is quietly running away. Finance loves this number. Operations hates looking at it. That's exactly why it belongs on the page.
How do you set the alert lines? Don't guess. Run two weeks, get your baseline, then draw the line. For UPH, look at your building's own average and variance over the past three months, and set the line 15% below baseline. And use two sets of lines — peak and off-peak. Volume doubles in peak season, UPH naturally dips; judging peak days by off-peak lines just teaches the floor to fake the numbers. Once set, let the ops manager adjust thresholds without asking the owner every time.
The three numbers that tell you about the trend
Inventory variance. How many SKUs don't match the book, and by how much money. Plenty of owners treat cycle counting as a cost center; I see it the opposite way — variance is the thermometer of warehouse health. My rule: alert when variance dollars exceed 0.5% of that day's shipped goods value, and go to full physical count after two straight weeks of alerts. It comes from the inventory adjustment table, and your adjustment reason codes have to be disciplined — otherwise you can't even tell "lost" from "miscounted."
Stockout cancellation rate. Orders canceled because there was no stock. Above 2%, the problem usually isn't the warehouse — it's replenishment rules or purchasing lead times. This metric gets pinned on operations unfairly, so split the cause: can't find it in the slot is the warehouse's problem; the system never had it is planning's problem. Mix them together and the number is useless.
Returns backlog days. Average days from RMA receipt to disposition. Past 5 days and customer service starts getting the angry calls; past 10 and your client's cash is sitting in your building. Most WMS standard reports don't include it — have IT compute it from the return order's status timestamps. Don't dismiss it: returns velocity is one of the first numbers clients ask about at 3PL contract renewal.
Layout matters too. One page, most critical number top-left — the ship-on-time rate. A red threshold line next to each number; cross it and it flags itself. No trend charts — those are for analysts. The owner needs one glance to answer 'is anything wrong today.' Three colors max: red, yellow, green. I've seen dashboards in seven colors. They look like Christmas trees and communicate zero.
Three traps that kill dashboard projects
Too many metrics. I've seen a cockpit with 40-plus indicators crammed on one page. The owner opened it twice and never again. A dashboard is for making judgments, not for research analysts. Six is the ceiling; anything else goes on a drill-down page for whoever cares.
Nobody trusts the numbers. Book says one thing, the floor says another — one "these numbers are wrong" from the ops manager and the whole page is dead. So before the dashboard goes live, get cycle counting solid and reason codes clean. Trusted data is the 1; everything after it is zeros.
Metrics divorced from incentives. Tie bonuses to UPH without adjusting for wave difficulty, and the team picking bulky items under the same UPH bar as the small-parcel team will lose half its veterans in three months. Metrics can be one-size-fits-all; incentives must be tuned per scenario. Otherwise the dashboard becomes a tool for driving good people out.
Finally, review the six numbers themselves every quarter. Any metric nobody clicked for three months gets cut; any metric the floor keeps asking 'what does this even mean' about gets its definition rewritten. A dashboard isn't set-and-forget — when the business changes, the numbers change with it. That Ontario client swapped 'returns backlog days' for 'customer complaint rate' in year two, because their returns process was fixed and complaints had become the new weak spot.
One thing you can do tomorrow: have IT build this as a one-pager tonight — no BI tool needed, a scheduled spreadsheet works. In your inbox by 8 a.m., every day, for 30 days. You'll find that most problems show up in these six numbers before they become incidents. Five minutes a morning buys you "knowing before it breaks." That's a trade worth making.
