Advisory Review Model
Running architecture review as a consulting service that improves designs rather than as an approval gate that permits them.
The approval-gate model produces predictable pathologies: teams present at 90% completion when nothing can change; the session is theatre; the reviewers carry accountability without context; and the queue becomes a lead-time problem that shows up in delivery metrics rather than in the review process's own.
The advisory model inverts the relationship. Teams come because the review makes their design better, which means it must actually do that — early, with reviewers who engage with the problem rather than checking conformance.
What makes it work in practice:
Early engagement, at 20% rather than 90%, where the design can still change.
Reviewers who have read the material beforehand, so the session is spent on judgement rather than presentation.
Review against published criteria — prioritised quality-attribute scenarios, principles, fitness functions — so the outcome is not determined by seniority.
A written record of decisions, actions, owners, and concerns deliberately accepted.
Fast turnaround, because a slow advisory service is avoided just as a gate is.
The small set of things that should still require approval: irreversible decisions, anything crossing a regulatory boundary, and commitments the organisation cannot easily exit. Keeping that list short is what makes the advisory model credible for everything else.