How to test a WMS in thirty days without disrupting the floor.

Week 1 — Stand it up and stock it

• Install and seed. Run the install on any machine and load the showcase world: eight client brands, three weeks of shipped history, live inbound, and an open order book that reconciles. Nothing here is painted on — every number came from the same engines the product runs on.
• Model one real client. Create one of your actual customers: their code, SLA, rate card, carriers and lot-code pattern. Import their SKU list by CSV and watch the dry-run report the rows it would reject before anything is written.
• Lay out one zone. Generate bins for a single aisle, print ZPL labels, and check the walk map matches how your people actually walk it.

Week 2 — Receive and store like it's real

• Receive a live inbound. Announce an ASN, book the dock, and receive it in parallel with your current system. Scan real GS1 barcodes off the cases; compare what one trigger pull fills in against what your team types today.
• Break something on purpose. Receive a short shipment, photograph a damaged pallet, put stock in the wrong bin. Then correct it — and see that the correction is a new event with a reason code, not an edit that erases the mistake.
• Try to violate a rule. Put ambient stock in the cold room, or a flammable in a general aisle. The refusal should name the SKU, the zone and what would satisfy it.

Week 3 — Run the floor

• Pick with a gun. Place orders against the stock you received, then work the queue on a phone or scanner gun: bin scan, product scan, count. Deliberately grab the wrong lot and confirm the system stops you.
• Batch a cart. Take the suggested cart and walk it. Compare the merged route against picking those orders separately — the suggestion states its own savings, so check the claim.
• Pack and ship. Box the order, weigh the cartons, and ship through whichever provider matches your setup: your aggregator's key, the watched-folder handoff for an on-premise system, or manual entry. Print the per-box packing slips.

Week 4 — Check the numbers and the paperwork

• Reconcile. Received minus shipped should equal on hand, orders should equal order fees, lines should equal pick fees. If a WMS cannot make those three statements true, its invoices are guesses.
• Price a month. Open the billable-events report for your pilot client and compare the rated total against what you would have invoiced by hand.
• Hand it to the floor. Give the pick and pack stations to whoever will actually use them, and watch without helping. Speed of the second day matters more than the first.

Questions worth asking any WMS vendor

• 1Where does my clients' data live, and what leaves the building? If the answer involves a multi-tenant cloud, ask what your clients' compliance teams will say.
• 2Can I explain a count to an auditor? Ask to see how the number got there. If stock is an editable field, the history is a story rather than a record.
• 3What happens to billable work between month-ends? Fees reconstructed at invoicing time are fees that get missed. Ask when the fee is recorded.
• 4How long does a new client take? If the answer is measured in development sprints, every unusual client is a project.
• 5What does the API not do? If the UI can do things the API cannot, integrations will eventually need screen-scraping or manual work.
• 6What happens when a carrier's API is down? There should be a path that does not stop the dock — manual entry, a file handoff, something.
• 7How do I leave? Ask for the export. The answer tells you how the relationship is expected to go.

What a pilot costs you

Time, mostly: a few hours a week from one operations lead and one floor supervisor. The install runs on hardware you already have, the demo world is one command, and nothing about the pilot touches your live system. If it does not earn its place by week four, you have lost four weeks of evenings and learned exactly what to demand from the next vendor.