it stopped working

Lesson 28: The 30-Minute Metrics Review That Stops the Drift

Lesson 28: The 30-Minute Metrics Review That Stops the Drift

Everything worked, and then quietly it stopped working.

That’s the pattern I’ve seen more than any other. Adherence hits ninety percent in week three. By week nine it’s at sixty and nobody has noticed, because there was no moment at which anybody was supposed to look. Incidents creep back one at a time, each with a reasonable explanation.

Nothing dramatic goes wrong. The system just slowly returns to how it was, and everyone involved remains convinced the improvement is still in place.

The Rule: Improvements Decay Unless Something Checks Them

This isn’t a discipline problem or a people problem. It’s the default. Attention moves to the next fire, and whatever isn’t being looked at drifts.

The counter-measure is small: thirty minutes, weekly, three numbers. The signal. The incident count. Adherence to whatever governance artifact you installed.

Three numbers. Not a status update. Not a review of everything the team did. Three numbers and the blockers stopping them from improving.

A Real Example: Eighty-Five Percent, Recovered in Two Weeks

After a set of changes landed well, I did what felt reasonable and stepped back. The team had it. Adding a recurring meeting to a team that had just improved felt like punishing them for succeeding.

Six weeks later adherence to the release checklist was under sixty percent. Two incidents had come back that the checklist was specifically designed to catch. The signal had gone flat and started to reverse.

Nobody had made a decision to stop. People got busy, a couple of releases went out under pressure with the checklist half-done, nothing bad happened either time, and the checklist became optional without anyone saying so.

We put in a thirty-minute weekly review with a fixed agenda and one requirement: every owner brings one remediation to the meeting. Not a status. A fix.

Adherence was back above eighty-five percent within two weeks. What did it wasn’t the measurement – it was that people knew, on Thursday, that they’d be sitting in a room on Monday where the number would be visible with their name next to it.

How To Do This Yourself

Fixed agenda. Guard it, because this meeting expands if you let it:

  • One minute: signal snapshot. The trend and the latest data point. No commentary.
  • Five minutes: incidents since last week. One line each. Not a post-mortem – that happens elsewhere.
  • Five minutes: adherence percentage and any notable exceptions.
  • Fifteen minutes: blockers and decisions. Every decision leaves with a named owner.
  • Two-line log of what was decided, circulated the same day.

Chase the owners before the next meeting, not during it. A review where people report on what they didn’t do becomes theatre within a month.

And end at thirty minutes even when the conversation is good. Especially then. The credibility of the meeting comes from it always being short.

The Honest Lesson

People hate meetings, and they’re mostly right to. I’ve sat through years of hour-long status calls where nothing was decided and everyone left with a slightly worse mood.

What I’ve found is that teams tolerate a short focused meeting for a few weeks and then start protecting it – but only if it removes blockers. The moment it becomes a place where people justify their numbers, it’s dead. You’ll know because attendance starts slipping and nobody says why.

The other correction I’d make to my younger self: don’t run this meeting yourself forever. If you’re the one holding it, the discipline is yours, not the team’s. Hand it to the flow owner after a month or two and see whether it survives. If it doesn’t, that’s information worth having.

Put thirty minutes in the calendar for next week with those four agenda items. Don’t add a fifth.

Next time, I’ll dig into “Document Wins and Failures, and Assign a 90-Day Follow-Up Owner” – see you there.