Software Architecture Implementation and Governance
This domain tests how architecture becomes working software while staying conformant: decision rights, reviews, automated enforcement, exceptions and architecture-level technical debt.
Core rules and concepts
- Implementation proceeds through a walking skeleton, architecture runway and enablers that make later feature work cheap rather than through a big-bang build.
- Governance defines decision rights explicitly: which decisions are central (platform standards, integration patterns, data ownership) and which are delegated to delivery teams.
- Conformance is enforced by a mix of peer review, architecture review boards and automated fitness functions in the pipeline: dependency and layering rules, API schema compatibility, policy as code.
- Exceptions are legitimate but must be recorded, time-boxed, owned and given an expiry and remediation path. Silent deviation is the failure mode, not deviation itself.
- Technical debt is tracked at architecture level too: duplicated capabilities, obsolete integration patterns, end-of-life platforms, and gaps in observability or operability.
Worked example
A team wants to call a partner API directly from browser code to hit a launch date, bypassing the integration layer. Governance records a decision with an exception: token brokering and rate limiting are added at the edge as compensating controls, the exception expires in ninety days, an owner is named, and a fitness function fails the build if partner endpoint patterns appear in front-end code. The reasoning: blocking the launch destroys business value, while an unrecorded bypass destroys architectural integrity; a time-boxed, automatically enforced exception preserves both and converts the rule from prose into a test.
Traps that cost marks
- Equating governance with a late approval meeting rather than continuous, principle-based steering.
- Granting exceptions with no expiry, owner or compensating control.
- Publishing architecture rules that no pipeline check enforces.
- Measuring governance by document volume, and ignoring operational readiness such as observability and runbooks as architectural concerns.