An indicator nobody acts on is a report, not an indicator
Homologation coverage, backlog, production bugs, team overload. Four numbers, one question each.
Most quality dashboards measure what is easy to count. The useful ones measure what someone is prepared to change their week over.
I keep four, and each of them has to answer a question that has an owner.
Homologation coverage
What share of what shipped was actually homologated? Not "tested at some point" — homologated, by someone accountable. When this falls, the question is not "who slipped": it is which demands were allowed through without capacity attached to them.
Backlog
Not its size. Its age. A backlog of forty items where nothing is older than two weeks is healthy. A backlog of twelve where four have been waiting since March is a prioritisation problem wearing a capacity costume.
Bugs in production
Counted by where they were introduced, not by who found them. A bug traced back to an ambiguous requirement and a bug traced back to a missed test case are two different problems, and only one of them is solved by testing more.
Team overload
Hours, but mostly concentration: how much of the queue depends on the two people who know the hardest flows. This is the number that predicts the other three, two months out.
The rule
If a number moves and nobody's week changes, it is a report. Stop publishing it and free the time — a dashboard with four live numbers beats one with twelve that everyone scrolls past.
