Stage 6. Opportunities, Solutions, Migration, and Delivery: phase summary

Phase E is where the enterprise stops describing its architecture and starts shaping delivery, grouping the consolidated gaps from Phases B through D into coherent solution paths. The four-fit test keeps groupings honest: business fit (a grouping serves one capability or value stream), information fit (it respects data authority and publication discipline), application and technology fit (components support each other without awkward cross-dependencies), and governance fit (exceptions to principles stay visible). Phase E is not vendor selection, not a disguised funding request, and not the final roadmap. In London this produced four cross-domain groups, starting with the case spine and evidence model that every later group depends on.

A real transition architecture passes five tests: a clear reason to exist, known compromises made visible, visible controls, a bounded exit path stated as testable architectural conditions rather than dates, and standalone value so the programme keeps its mandate and has a safe fallback. A work package is an architecturally meaningful unit of change with a scope, dependencies, and a transition-state assignment, not a sprint, project, release, or budget line. London's three transition states (evidence foundation, then publication and modelling discipline, then full operating capability) each deliver recognisable value, and the work packages WP1 through WP4 carry explicit dependency chains between them.

Phase F finalises the roadmap and produces the Implementation and Migration Plan, which translates architecture order into delivery terms for programme management. Sequencing balances seven criteria: dependency (the strongest, because violating a genuine one causes failure), value, risk, readiness, cost and resource, regulatory and contractual, and stakeholder impact. The recurring failure is decorative sequencing, where bars are arranged to look balanced across quarters rather than to reflect dependency, readiness, or risk. London's sequence can answer the review questions for every step: data-authority codification comes before publication migration because publishing from unclear authority creates compliance risk, and Ofgem transparency deadlines are fixed inputs.

The ADM is iterative by design (within-phase, cross-phase backward, Architecture Development, Transition Planning, Architecture Governance, and multi-level patterns from C220 Part 3), and governed looping (explicit board decision, defined scope, impact assessment, time bounds) strengthens the architecture while uncontrolled looping wastes it. Partitioning by capability and value stream manages complexity but must keep cross-cutting dependencies visible in the repository; the three landscape levels (strategic, segment, capability) form a chain of consistency that also feeds portfolio decisions (G238). Agile and architecture coexist through explicit decision rights, the G210 architecture runway, G20F tailoring, and the G188 bridge to project management: boundary decisions, authority rules, and transition logic stay stable while detail design within guardrails iterates.

Readiness assessment is a sequencing input, not a late training exercise. The C220 Part 3 factors (vision, desire, need, business case, funding, sponsorship, governance, accountability, IT capacity, enterprise capacity) are each scored on readiness rating and urgency of improvement; low readiness with high urgency defers a work package or slows the pace. Genuine readiness happens before commitment, has authority to change the plan, and covers every dimension. London's synthesis ties it together: WP1 starts first on its strong readiness and regulatory driver, and the three evidence gates test architectural conditions (evidence quality, information authority, assurance integration, G21H compliance conformance) so the board can hold, revise, or proceed on evidence rather than on whether a milestone date was met.

• Letting procurement or product boundaries replace architecture boundaries by starting vendor conversations before Phase E grouping is stable.
• Jumping from baseline to target without designing the intermediate states, so temporary compromises drift into permanent fixtures nobody documented.
• Arranging the roadmap to look balanced across quarters instead of reflecting genuine dependency, readiness, and risk, and deferring the highest-risk work to last.
• Treating readiness as a training plan bolted on at the end, assessing only technology readiness, or running the assessment after the plan is already fixed.