Failure Note - Floor gate

No release gate means no production decision.

When 'ready' was never written down, the goalposts move until the sponsorship expires.

By Jay Sharma · September 20, 2026

The claim

A team that never wrote down what "ready" means will keep moving the bar. The pilot does not get killed and does not ship. It drifts until the sponsorship expires.

How it fails

The launch criterion is a feeling: "when it is solid." Each stakeholder carries a private bar, and the disagreement surfaces as delay rather than decision. After every demo, someone raises the threshold - not because the system got worse, but because the threshold was never fixed, and raising it is the only lever anyone has.

Warning signs

  • The go-live criterion is a vibe rather than a document.
  • The bar has changed after at least one review, without new evidence.
  • Nobody can state the minimum acceptable performance, the allowed blast radius, and the rollback plan for a first release.
  • Every review ends with "one more improvement" rather than a decision.

What it costs

The pilot dies of drift instead of decision. The team loses a year, the sponsor loses patience, and the next proposal inherits the skepticism this one earned.

What would change the answer

A written release gate agreed while the stakes are low: minimum thresholds, exposure limits, rollback criteria, and the named person who can say go. If that document exists and everyone signed it, this note does not apply to you.

Illustrative composite - method walkthrough, not a client result

An underwriting assistant watched its accuracy bar move from 85% to 92% to "let us see one more quarter" across three reviews. No new evidence drove the changes. Funding lapsed in the fourth quarter, with the system two fixes from a bar that did not exist. The gate that failed was release: there was no production decision because there was no production standard.

Where this maps

This failure is a Floor gate failure. The PROOF Scorecard tests this gate in about four minutes; the full review proves it with evidence.

Related notes

Bring one pilot and one decision.

Thirty minutes on one stuck pilot: the decision being avoided, the five production gates, and the first one that lacks evidence. If a full review is not useful, I will say so.

Book a private pilot triage