Govern
Governance is one loop and one habit: local improvements become the new standard cheaply, and plays retire in the open. Everything heavier than that is theatre.
Play: local improvement to new standard
This is the play the whole system exists to serve.
When: a team has varied a play locally, the variation has survived at least three uses, and it outperforms the standard on a signal the play already names.
The action: the team proposes the change with the evidence, not the argument. The proposal path is the same lightweight mechanism every time, and it costs less than the improvement is worth. The play owner either adopts, adapts, or declines with reasons that reference the evidence. The play's last-tested date and status update either way.
Signals: the standard actually shifts, meaning other teams' behaviour changes, not just the document. Proposals come from teams you did not prompt. The gap between the official play and what teams really do stays small.
When not to: when the variation only worked because of conditions the proposing team has and others do not. Then it is a fork, not an improvement, and the honest answer is a second play with different entry conditions.
The failure this prevents: the moment updating a play costs more than ignoring it, people route around the playbook and it forks into an official version and a real one. That is the espoused playbook arriving by a different door.
Play: retire in the open
When: a play's last-tested date is old, or its signals have stopped firing, or the situation it addressed no longer occurs.
The action: mark it retired, say why, and leave it visible. Retirement is a first-class outcome, not a failure. A retired play with its reasons attached teaches almost as much as a live one.
Signals: the collection has retired plays in it. A playbook that has never retired anything has no governance, whatever its process documents say.
When not to: when the play is untested rather than dead. An untested play gets a status flag and a test, not a funeral.
Antipattern: the museum piece
Written once, never revised against evidence, no feedback loop. Stale plays are worse than no plays because they carry false authority. The tell: every play has the same date on it.
Antipattern: governance theatre
Heavyweight process for a lightweight artefact. Review boards, approval chains, quarterly cadences for a document that should change the week the evidence arrives.
The proportionate mechanism is small: every play carries an owner, a status, and a last-tested date. Anyone can propose a change by evidence. The change path is cheaper than the workaround. That is the whole system, and it borrows deliberately from RFC and ADR practice: decisions carried with the artefact, not in a committee's minutes.
Licensed under CC BY-SA 4.0.