How this is graded
Every control carries an evidence tier. The tier describes the strength of the study behind the claim, not how much we like the claim.
| I | controlled experiment or large-N study with a comparison group |
| II | observational study on production data, no control group |
| III | single-org case study or engineering report, with numbers |
| IV | argument or experience report — not measured |
Tier IV entries are allowed. A great deal of good practice has never been measured, and pretending otherwise would be its own kind of dishonesty. What a tier IV entry must do is say what measurement would move it up. If it cannot answer that, it is not a control — it is a preference.
Field notes are not evidence
Some controls carry short accounts of how a thing goes wrong in practice. They are there because an abstract control is hard to recognise in your own system. They carry no tier, they are not citations, and the build fails if one is cited as evidence.
Retractions are visible
Evidence changes. When a claim here is downgraded or withdrawn, the page says so rather than quietly editing itself. A catalog that has never retracted anything has not been running long enough to trust.
This is not a maturity model
No levels, no score out of five, nothing to certify. A position and a bearing, never a grade.