How do I diagnose a failed WMS implementation?

The diagnostic question most founders ask after a failed implementation is what the implementation team did wrong. That question rarely produces a useful answer. Implementation teams don't usually fail the way the question implies — they fail to recover from a fit problem that was already locked in before they started.

Most WMS implementations that fail trace back to one of three structural causes, and all three sit in selection rather than implementation.

The first is that the wrong system was selected. Not in the sense of a poor product — in the sense of a system that fit the operation as it was described in selection but not the operation as it actually runs. The workarounds, edge cases, and process realities that make this specific operation work weren't surfaced as requirements because proximity blindness had made them invisible to the founder before the evaluation began. Implementation became the moment when the un-surfaced operation met a system selected without it.

The second is that the operational picture the system was selected against was inaccurate — which is distinct from picking the wrong system, because the right system selected against an inaccurate picture fails the same way. What broke wasn't the product — it was the assumptions about throughput, exception rates, integration realities, and process maturity that sat underneath the selection. Founders don't see this clearly during selection because every input — requirements, demo scenarios, reference checks — runs through the same description-based picture.

The third is that the vendor over-promised on capabilities that don't exist as demoed, or exist only with services-team work the contract didn't price. Strategic omission during demos is near-universal in the WMS market — not as dishonesty but as the absence of information the buyer didn't know to ask about. Roadmap features framed as if shipped. Configurations that work in a vendor-controlled sandbox but break in a fresh environment. Integrations that work in principle but fail in the specific ERP version you actually run. From inside a buyer-only frame, those gaps look like the system performing as demoed. From the inside of the vendor preparation that produced the demo, they're the predictable result of a sales motion calibrated to satisfy what buyers know to ask about.

The implementation team inherited the consequence. Whatever they did or didn't do, they were solving for a fit problem the selection process couldn't see — at the point in the project where changing course costs the most. The diagnostic question that produces a useful answer isn't what implementation got wrong. It's which selection-stage gap is now visible in the failure.