Keep the alternatives that lost next to the diagram.

Question: stay, move a workload, buy a managed service, split a boundary, or postpone. Owner: usually an architect or engineering lead. Stakeholders: security, operations, and the budget owner.

A diagram that shipped without the rejected paths is a poster, not a record.

Reproduced workload characteristics beat vendor benchmarks. Teams disagree about lock-in, cost, and who gets paged. Those disagreements belong on the record.

Known unknowns: unreproduced numbers, hiring that has not landed, and an exit path nobody has tried. Material risks: blast radius and a constraint nobody can reopen.

Decision Hinges: time to value with this team, exit, and operational load. Required approvals: Decision Owner plus security when blast radius is material.

What could reverse the Verdict: a hiring slip, a failed proof, or a cost that only appears in production. Outcome review: after the first production incident that the new boundary was supposed to contain.

Champion defends the preferred architecture. Challenger attacks operations and exit. Strategist may propose a thinner path. Inspector marks unreproduced benchmarks. Judge scores time to value with this team.

Modern is not a criterion. A proof of concept is not production.

Champion and Challenger test the case, Inspector checks the evidence, and Judge records the Verdict for the Decision Owner.

Keep the current boundary, split a service, or buy a managed substitute. Write the exit. If the Verdict favors split and the Human Decision stays because hiring will not land this half, both remain on the record.