Architecture Principles

How to write principles with rationale and implications, a sample set for a cloud-native enterprise, and how to make principles enforceable rather than wall posters.

Format

TOGAF recommends each principle have a name, statement, rationale and implications. Implications are what make a principle useful: they tell teams what to do differently.

Sample principles

Name Statement Rationale Implications
Business capability alignment Solutions are organized around business capabilities Stable boundaries, clear ownership Service and team boundaries follow capability map; no shared databases across capabilities
Buy for commodity, build for differentiation Use SaaS/COTS where capability is not a differentiator Focus engineering on competitive advantage Build decisions require a differentiation case
API-first Every capability is accessible through documented, versioned APIs or events Reuse, composability OpenAPI/AsyncAPI specs required; catalogued in API portal
Data is an asset with an owner Each data domain has an accountable owner and classification Quality, privacy, compliance Data catalog entries, classification tags, retention rules mandatory
Secure by design and default Security controls are built in from design Lower risk and cost Threat model for Tier 0/1; secure defaults in templates
Cloud-native, managed first Prefer managed services over self-managed infrastructure Lower operational load Self-managed requires justification
Design for failure Assume components fail Resilience Timeouts, retries, redundancy, DR tier declared
Observable by default Every service emits standard telemetry Faster incident resolution OpenTelemetry instrumentation in starters
Automate everything Infrastructure and deployments are code Repeatability, auditability No manual production changes
Cost-aware Cost is an architectural quality attribute Sustainable growth Unit cost tracked; cost reviewed in design

Making principles real

  • Few (8–12), memorable, with clear implications.
  • Referenced in ADRs and reviews ("this decision follows principle X / deviates because...").
  • Encoded into templates and automated checks where possible.
  • Reviewed annually.

Interview questions and answers

Two principles conflict, for example "buy for commodity" and "data is owned and controlled". How do you resolve it?

Make the trade-off explicit for the specific case: evaluate the SaaS vendor's data controls (export, residency, encryption, APIs). If acceptable, buy; if not, consider alternatives. Record the reasoning in an ADR referencing both principles.