The Procurement Window
An insurer is replacing its policy administration platform. The architecture team has produced a requirements catalog, a business service model, a data entity model, an application communication diagram and a candidate roadmap containing seven work packages. It has also just signed a Statement of Architecture Work - three months after work began - with a sponsor who has since been replaced. The team is nominally in Phase E. Procurement faces a hard window: long-lead infrastructure must be ordered in three weeks or the programme slips a full year. The new sponsor has never seen the Architecture Vision, and the CTO wants a decision on the database platform this week so that licence discounts can be negotiated.
1. The team proposes to move straight into Phase F next week to catch the procurement window. Set out what is missing against the ADM, and what a defensible minimum would be before money is committed.
Read the worked answer
What is missing. Phase A was never properly closed: the Statement of Architecture Work was signed late and by a sponsor who has left, so the work is not currently authorised; the Architecture Vision has not been accepted by the new sponsor; the stakeholder map and capability assessment are stale. On the other side, Phase E is incomplete: there is a candidate roadmap of seven work packages but no agreed consolidated gaps and dependencies view, and no agreed transition architectures saying which packages deliver which intermediate state. Phase F is supposed to take those as inputs and turn them into an Implementation and Migration Plan with priorities, dependencies, costs and business value. Entering F on a candidate roadmap means the plan has no traceable basis.
A defensible minimum. (1) A one-to-two-week re-baseline of Phase A with the new sponsor: Vision restated and accepted, Statement of Architecture Work re-signed, stakeholders refreshed. (2) Confirmation of the Phase E work packages and at least one transition architecture that the delivery organisation accepts. (3) For the long-lead items only, a narrow Phase F slice - dependencies, costs, lead times and an explicit statement that the technology may change as the cycle completes. (4) The decision and its accepted risk recorded in the Governance Log.
The judgement to make is between irreversible and reversible commitments. Ordering long-lead infrastructure with a documented risk can be justified; using the deadline to authorise the whole programme cannot.
2. The CTO wants the database platform decided this week. Where does that decision belong in the ADM, and how do you handle it without stalling the cycle?
Read the worked answer
The database platform is Technology Architecture, so the substantive decision belongs in Phase D. Deciding it before the business and information systems architectures are settled is premature: requirements that emerge in B and C are precisely what should discriminate between candidate platforms, and a decision taken now may lock in something the data architecture later contradicts.
That does not mean refusing the CTO. The mechanisms available are:
- Route the request through Requirements Management, which sits at the centre of the ADM and is the legitimate channel for a requirement raised out of phase order.
- Run a focused Phase D iteration scoped to the data platform, using the business and data requirements that already exist, rather than waiting for the full cycle to arrive at D.
- Structure the output as an options-based selection with defined exit criteria and contract terms permitting change, so the decision is a direction rather than an irreversible lock-in.
- Record the chosen candidate standard in the Standards Information Base and the rationale as an architecture decision in the Governance Log.
- Feed the selection into Phase E so it is consolidated with the work packages.
What must not happen is that an executive decision bypasses the requirements flow. If it does, the architecture loses traceability, Phase G compliance reviews have nothing to test against, and the platform becomes a constraint nobody can later justify.
3. The Head of Architecture proposes to skip Phase G for this programme because 'the PMO already does assurance'. What is Phase G for, and what would be lost?
Read the worked answer
Phase G is Implementation Governance. It is where the architecture stops being a document and starts constraining delivery: architecture contracts between the architecture function and the delivery teams, compliance reviews at defined checkpoints, handling of change requests raised during implementation, and reporting of exceptions back to the Architecture Board.
A PMO assures cost, schedule and delivery method. It does not test conformance to the target architecture, it cannot grant a dispensation, and it does not own the architecture contract. If Phase G is skipped, three things are lost. First, the target architecture becomes advisory - deviations surface at go-live rather than at a checkpoint where they can still be corrected. Second, there is no record of what was approved, by whom, and on what risk basis, which matters here because the programme has already departed from a signed contract. Third, Phase H has nothing to act on: change management works from the governance record of implemented architectures.
The right answer is integration, not duplication - place the architecture compliance review at the PMO's existing gate points and reuse its reporting, but keep the conformance judgement and the architecture contract with the architecture function.