ADR-020: 7-Phase to 5-Stage Lifecycle Migration
| Field | Value |
|---|---|
| Status | Accepted |
| Date | 2026-05-02 |
| Decision Makers | HITL, Cloud Architect, Product Owner |
| Scope | adlc-lean5s3t-standardize-2026-05-02 |
Context
Three lifecycle models coexisted in the ADLC framework:
| Model | Phases | Origin |
|---|---|---|
| ADLC 7-phase | Think → Plan → Build → Review → Test → Ship → Reflect | ADR-019 (2026-04-28) |
| Command Center PM 6-phase | Discover → Specify → Plan → Build → Ceremony → Iterate | product-management.md |
| CloudOps Infra 6-phase | Discover → Design → Validate → Deploy → Operate → Optimize | cloudops-infrastructure-lifecycle.md |
Translation overhead between three models consumed an estimated 12–210 hours per year in CxO meetings and onboarding sessions. Engineers and stakeholders had to mentally map between models when reading framework documentation.
On 2026-05-02, HITL confirmed adoption of the 5-stage canonical (5-canonical-yes) using the Anthropic Claude Code workflow image as the normative SSOT. The 5-stage model aligns with three industry frameworks:
- Anthropic Claude Code: Discover · Design · Build · Deploy · Support & Scale
- McKinsey Rewired (Exhibit 5 Ds): Discover · Define · Design · Develop · Deploy
- ITIL v4: Value chain stages map cleanly to Discover/Design/Build/Deploy/Support
Decision
Adopt the 5-stage canonical lifecycle as the single standard for ADLC documentation, CSV data, and TypeScript types. Replace the 7-phase ADLC canonical as the primary vocabulary. Retain the 7-phase model as a legacy mapping in data-quality-standards.md for historical coordination log compatibility.
Operative 7→5 Mapping Table (canonical SSOT for all migration scripts)
| 7-phase | 5-stage | Sub-step | Autonomy Tier | Rationale |
|---|---|---|---|---|
| Think | Discover | (single) | [A] | Both = problem understanding and assumption surfacing |
| Plan | Design | (single) | [A] | Both = solution design, ADR drafting, scope binding |
| Build | Build | code | [A] | Direct match — implementation |
| Review | Build | review | [A-gated] | Code review is a Build sub-step (PR review before merge) |
| Test | Deploy | test | [A-gated] | Test = pre-deploy verification gate |
| Ship | Deploy | apply/release | [HITL] | Deploy includes the HITL-executed release moment |
| Reflect | Support & Scale | monitor | [A-readonly] | Both = post-deploy operations, MELT/DORA tracking |
| Reflect | Support & Scale | remediate | [HITL-decide] | RCA → HITL approves remediation actions |
Sub-step autonomy tiers are preserved from ADR-019. The 5-canonical collapses 7 phase labels into 5 stage labels without losing governance granularity — sub-steps within each stage carry the original autonomy tier.
Python Migration Mapping Dict
Used in Wave 2.5+2.6 CSV migration script (scope_id: adlc-lean5s3t-standardize-2026-05-02):
MAPPING = {
"Think": "Discover", "Plan": "Design",
"Build": "Build", "Review": "Build",
"Test": "Deploy", "Ship": "Deploy",
"Reflect": "Support & Scale",
"THINK": "Discover", "PLAN": "Design",
"BUILD": "Build", "REVIEW": "Build",
"TEST": "Deploy", "SHIP": "Deploy",
"REFLECT": "Support & Scale",
"DEPLOY": "Deploy", "Deploy": "Deploy",
"MONITOR": "Support & Scale", "Monitor": "Support & Scale",
"OPERATE": "Support & Scale", "Operate": "Support & Scale",
}
Implementation Files (Wave 2.5+2.6)
The following files were updated as part of this ADR's implementation:
| File | Change |
|---|---|
docs/src/data/projects/types.ts | ADLCPhase union updated to 5-stage |
.claude/rules/quality/data-quality-standards.md | Phase canon section replaced (7→5), legacy mapping added |
docs/static/data/jira/adlc.csv | adlc_phase column migrated |
docs/static/data/jira/security.csv | adlc_phase column migrated |
docs/static/data/jira/cloudops-runbooks.csv | adlc_phase column migrated |
docs/static/data/jira/devops-terraform.csv | adlc_phase column migrated |
docs/static/data/jira/finops-focus.csv | adlc_phase column migrated |
docs/static/data/jira/platform-idp.csv | adlc_phase column migrated |
docs/static/data/jira/itsm.csv | adlc_phase column migrated |
docs/static/data/jira/cloudops.csv | adlc_phase column migrated |
docs/docs/getting-started/workflow.md | Phase references aligned to 5-canonical |
docs/docs/command-center/product-management.md | 6-phase PM → 5-canonical |
docs/docs/golden-paths/cloudops-infrastructure-lifecycle.md | 6-phase Infra → 5-canonical |
Consequences
Positive
- One vocabulary: CxO, engineers, and AI agents share the same 5-stage language. No translation table in meetings.
- Industry alignment: McKinsey Rewired, Anthropic Claude Code, and ITIL v4 all map cleanly. New hires onboard faster.
- CxO speakability: "We are in the Build stage" is boardroom-ready. "We are in the Review phase" required explanation.
- Stakeholder savings: Estimated 12–210 hours per year eliminated from vocabulary translation overhead.
- Sub-step clarity: Review and Test are explicit sub-steps of Build and Deploy respectively — not separate stages. Autonomy tiers are preserved with more precision.
Negative
- One-time migration cost: 9 CSV files + 3 documentation files + TypeScript union updated in this PR.
- Historical coordination logs: 113+ existing logs in
tmp/retain 7-phase tags as immutable evidence. The legacy mapping table indata-quality-standards.mdresolves them during rendering. - Tooling updates:
task ssot:validate(XC-319) must be extended to validate against the 5-set canon after this ADR lands.
Alternatives Considered
Option A: Keep 7-phase ADLC canonical, translate in documentation only
Rejected: Translation overhead is the problem. Adding more translation layers compounds it.
Option B: Adopt McKinsey 5 Ds (Discover/Define/Design/Develop/Deploy) directly
Rejected: HITL confirmed Anthropic Claude Code image as SSOT. "Define" and "Develop" are less precise than "Design" and "Build" for an AI-agent engineering context.
Option C: Keep three separate models per domain (PM/Infra/ADLC)
Rejected: Three models = three sets of meeting vocabulary + three CSV phase enums + three documentation structures. Complexity compounds over time.
References
- Anthropic Claude Code workflow image (HITL-supplied 2026-05-02)
- McKinsey Rewired, Exhibit 5 Ds (Digital Factory operating model)
- ITIL v4 Service Value Chain
- Claude Code Documentation
- ADR-019: Autonomy Tier Spectrum — sub-step tier labels preserved in this ADR's mapping table
data-quality-standards.mdv2.0.0 — Phase Canon section updated with 5-set canonical and legacy 7-phase mapping
Origin: ADLC Framework Wave 2.5+2.6 standardization (scope_id: adlc-lean5s3t-standardize-2026-05-02). Elevated to ADR because this decision governs all lifecycle references across ADLC consumer projects.