Here's my verdict up front: a WMS POC evaluation is not a demo. It is three days running your own real orders, inventory, and exceptions in your own warehouse. If it doesn't survive those three days, don't sign. A demo that looks great tells you almost nothing about what happens on go-live night.

In the fall of 2024, I got called in to rescue a go-live at a home-goods 3PL outside Los Angeles. The warehouse ran roughly 800 orders a day from about 20,000 square feet, up to 1,500 in peak. The demo had looked great: clean picking screens, one-click wave creation. Then week one blew up. They had a category of oversized furniture that needed two people to carry and a separate outbound flow, and the demo had never once run that flow. Pickers couldn't find an entry point in the system, so they worked off paper tickets and rekeyed everything afterward, adding two to three hours a day. The owner told me he had actually asked about oversized items on demo day. The answer was "supported." He didn't push further. That one word cost three weeks of custom work, during which half the orders ran on paper plus rekeying. Three orders shipped wrong. One freight claim.

Why the demo always looks great

A demo isn't fake. It's curated. The vendor's dataset is polished: tidy locations, clean batches, nothing but standard orders. Your warehouse isn't like that. Yours has long-tail SKUs, damaged stock, customers changing addresses at the last minute, and a mixed batch that arrived yesterday. None of that shows up in a demo.

There is a sneakier problem, too: the person driving the demo works for the vendor. They know the system better than their own house — which button to press, which field to skip. Hand the same flow to your picker on their first day and it takes twice as long. That's exactly what a POC is supposed to measure: can your people, not their most practiced person, get the work done? A flow that still runs smoothly after you swap in your own staff is the only kind that counts.

How I run a three-day WMS POC evaluation

I schedule every POC the same way: three days, no fewer.

Day one: load real data, receive and put away. No vendor sample data. Bring in last week's actual receiving documents, SKU master data, and location map. Watch two things. First, can items with batch numbers and expiry dates import correctly on the first try, and does the system flag errors or swallow them? Second, do the putaway suggestions make sense, or is it sending heavy goods to high shelves and fast movers to a far corner? By the end of day one, system stock has to match your physical count. If it doesn't, stop and find out why instead of pushing ahead.

Day two: run the full outbound chain on yesterday's real orders — wave, pick, verify, pack, weigh, ship, all of it. And deliberately inject three kinds of exception orders: a short pick that forces a split, an address change mid-day, and a multi-piece order where one item won't scan. How exceptions get handled is the main event of a POC. Anyone can run the happy path; exceptions decide what time you go home after go-live. For every exception, write down what the system prompted, how many steps it took, whether the action could be undone, and whether inventory came back right.

Day three: cycle counting, returns, and reports, plus a network-outage drill. Returns receiving, damaged-goods transfer, cycle counts — demos barely mention these, but you'll do them daily after go-live. For reports I only care about three: the open-order list, inventory accuracy, and picker productivity. If any won't come out, or the numbers don't tie, that's a fail. The outage drill is simple: kill the network on the picking devices for ten minutes. Can they keep scanning offline, and does any data go missing when it comes back? I've seen this exact failure three times — every one discovered during peak season, after the network flickered and half an hour of scans vanished.

When the three days are done, write a go-live-night checklist: which flows passed, which only work through workarounds, which need custom work, and how long that will take. If more than three flows depend on workarounds, don't sign the contract yet. A workaround isn't forbidden, but each one is a landmine after go-live, and too many will go off eventually.

Demo sign-off vs. POC sign-off

Check Demo sign-off POC sign-off
Data Vendor's polished sample data Your real receiving documents and orders from last week
Who operates it The vendor's sales consultant Your pickers and inventory controllers
Exceptions Barely shown Three kinds deliberately injected
Pass standard Looks good, features present Stock ties out, exceptions handled, reports come out
Problems found Almost none Should surface three to five things that must change

Now the math. A serious WMS POC evaluation takes about two of your people for three days, plus roughly a week of vendor preparation. Compare that with one rework cycle after go-live — process changes plus retraining — easily two to four weeks for a mid-sized warehouse, and if it collides with peak season, the lost orders and customers don't fit in any spreadsheet. A POC costs a certain three days. It saves an uncertain month. I walk every client through this math, because the cost of a POC is visible and the cost of a failed go-live is not — and people always gamble on the cost they can't see.

Here's the contrarian point: if your POC runs too smoothly, be suspicious. Three days with no exception triggered and no flow stuck usually doesn't mean the system is perfect. It means your test orders were too polite. A good POC should find problems — three to five things that must change — or it hasn't earned its keep. A POC that finds nothing is the same as no POC at all. I've said this to clients many times. They nod, then still test the demo's happy path and sign.

In early 2025, an auto-parts warehouse in Texas learned this the hard way. They represented a dozen-plus brands with around 40,000 SKUs, and their biggest fear was switching between clients: the same picker on client A in the morning and client B in the afternoon, with different locations and rules. The demo tested one client. It was smooth. In the POC I insisted they run two clients in rotation, and it broke on day two: switching clients forced a fresh login, two minutes each, a dozen-plus switches a day — half an hour gone just logging in. Worse, picking under the wrong client didn't trigger any warning — the goods went straight into the wrong wave. If the POC hadn't caught it, the first week after go-live would have shipped wrong orders. The vendor changed the switching logic: one-click switch, no re-login, forced verification on every switch. I've seen this same failure three times now, always at multi-client warehouses where the demo tested a single client.

The five failure points I see most

First, testing only the standard flow and skipping exceptions. Exceptions are your daily reality after go-live; the standard flow is only the demo's.

Second, using sample data. Sample data has no dirty data. Your warehouse is full of dirty data. A POC must be tested on dirty data.

Third, letting the vendor's people operate the system. If your team never touches it, the POC measures the sales consultant's speed, not your team's learning curve.

Fourth, leaving reports until last. If the reports don't come out, you're flying blind after go-live. Look at them on the morning of day three, not at the last minute.

Fifth, accepting verbal promises for custom work. "Supported," "we can build it," "phase two" — unless it's written in a contract appendix with a schedule and a price, treat it as not supported. A verbal promise is the cheapest thing in the room on go-live night.

One thing you can do tomorrow morning: print the ten hardest orders from last week — the short pick, the address change, the unscannable item, the oversized piece that needs two people. Take them into your POC conversation and tell the vendor: if these ten don't run, we don't sign. Those ten orders are worth more than a hundred pages of feature checklists from demo day.

By Lin Xiao, WMS Implementation Consultant