architecture-review-boardlisted

# Architecture review board An ARB exists to make the expensive, hard-to-reverse decisions once, with the right people present, and to write down why. Its failure modes are a matched pair: too loose and it stamps whatever arrives, too heavy and it becomes a tollbooth every team learns to route around. A board worth keeping reviews only what matters, decides with real authority, and leaves a record the next architect can trust. ## Method 1. **Set a submission bar keyed to blast radius.** Require review when a proposal adds a new datastore or language, crosses team boundaries, changes a shared contract, or opens a security or compliance surface. Apply the one-way-door test: reversible, two-way-door choices stay with the team. Everything else is noise the board should decline at intake. 2. **Demand a written proposal, not a calendar hold.** The submission is a design doc or ADR draft stating the problem, the options weighed, the recommendation, and the trade-off accepted. A board that reviews slides reviews charisma. Reject anything with a single option: one option is a decision already made, dressed as a question. 3. **Run a shepherd pre-review before spending the board.** One reviewer reads the proposal cold, confirms it clears the bar and answers the obvious questions, and either advances it or returns it. This keeps the full board out of triage and the author out of a public ambush over a missing section. 4. **Judge through fixed lenses so