The pilot missed its metric. We wrote it up anyway.

A pilot in a distribution business was built against a stock-out figure that we and the client both believed was computable. It was not, in the end, because the inventory snapshots were taken at a cadence that made the specific question unanswerable at the required precision. We could produce a number. We could not produce one anybody should sign.

So the pilot was reported as not clearing its metric, which is what the contract says happens. The client had paid for two weeks of discovery and a pilot, not for a rollout, and they stopped knowing something they had not known before: that a system change they had been deferring was blocking a class of measurement entirely.

Why this is the product working

A vendor with a rollout to protect finds a way to declare the pilot a success, because the alternative is losing the deal. That is not dishonesty, it is incentive, and it is exactly why the stages are priced and gated the way they are.

We would rather lose a rollout in month three than defend an unprovable number for two years. Our references answer the phone because of engagements like this one, not despite them.