Architecture Governance and Review Boards
Designing governance that enables delivery: decision rights, architecture review boards, guardrails, automated compliance, exceptions and measuring governance effectiveness.
Purpose
Architecture governance ensures that significant decisions align with strategy, principles and standards, manage risk, and avoid duplication, without slowing teams down unnecessarily.
Decision rights: who decides what
| Decision type | Decided by | Mechanism |
|---|---|---|
| Within a team's service (libraries, internal design) | Team | ADR in repo |
| Cross-team impact (new shared API, data ownership) | Domain/solution architect with teams | Design review, ADR |
| New technology/platform, deviation from standards | Architecture review board / technology council | Review + exception record |
| Enterprise-wide strategy (cloud provider, ERP) | CTO/CIO with exec stakeholders | Investment/strategy board |
Architecture review board (ARB) that works
- Clear triggers: only reviews meeting criteria (new technology, Tier 0/1 systems, cross-domain data, security-sensitive, cost > threshold).
- Lightweight submission: a short design doc template (context, decision, options, NFRs, risks) and ADRs.
- Fast cadence: weekly slots, decisions within days, asynchronous reviews for simple cases.
- Advisory and enabling: reviewers help teams solve problems; most outcomes are "approved with recommendations".
- Exceptions: time-boxed, recorded, with remediation plans and owners.
- Transparency: published decisions and rationale.
flowchart LR
T[Team design] --> Q{Meets review<br/>triggers?}
Q -- no --> ADR[Team ADR in repo<br/>automated checks]
Q -- yes --> R[Review: async or ARB slot]
R --> D{Decision}
D -- approved --> GO[Proceed]
D -- approved with conditions --> GO
D -- exception granted --> EXC[Exception register<br/>expiry + remediation]
D -- rejected --> RW[Rework with guidance]Automate guardrails
Shift governance into tooling: policy as code in cloud landing zones, CI checks (dependency allowlists, security scans, API linting), golden-path templates, fitness functions (ArchUnit rules, latency budgets in tests). Human review is reserved for genuinely novel or risky decisions.
Measuring governance
Lead time for reviews, percentage of decisions automated, exceptions open/expired, reuse of standards, incidents linked to non-compliance, and team satisfaction with the process.
Interview questions and answers
Teams complain the ARB slows them down. What do you change?
Narrow the triggers, publish templates and standards so most decisions do not need review, move to asynchronous reviews with SLAs, embed architects in teams early instead of gating late, automate checks, and measure review lead time publicly.
How do you handle a team that wants a technology not on the approved list?
Assess the need (what problem, why existing options fail), evaluate operational impact (support, skills, security), allow a time-boxed pilot as a recorded exception, and decide whether to adopt it as a standard, keep it as a sanctioned exception, or reject with alternatives.