The Blueprint

How Teams Build

The delivery standard for NHS Wales digital: discovery to live with kill gates, product not programme, published DORA metrics, golden paths.

Architecture says what gets built; the delivery standard says how. Wales does not need to invent one. The Centre for Digital Public Services already publishes a Digital Service Standard for Wales, and the NHS service standard exists alongside it. The blueprint adopts them, adds the health-specific discipline, and — the part that matters — attaches consequences.

The standard, in one screen

Every digital service built for NHS Wales, by the standards body or a health board team:

  1. Starts with users. Discovery before commitment. No business case without named clinicians and patients who have been observed doing the work the service touches.
  2. Moves discovery → alpha → beta → live, with kill gates. Each phase ends in an assessment by assessors outside the delivering organisation. Failing an assessment stops funding. The option to stop is the point: the current portfolio has nine programmes and no stop list; the standard makes stopping a normal, survivable event.
  3. Ships as a product, not a programme. A named product owner with decision rights; a team that stays with the service in live; roadmaps in public. Programmes end; products are run.
  4. Publishes its delivery metrics. The four DORA measures — deployment frequency, lead time, change-failure rate, time to restore — published per team, per quarter, on the national dashboard. Not as surveillance: as the profession’s own scoreboard. Teams that ship weekly have nothing to fear from a chart.
  5. Builds on the golden path. The standards body maintains platform tooling — pipelines, environments, observability, security scanning — so a new team ships its first change in days. Divergence is allowed and documented, never accidental.
  6. Meets the invariants. WCAG 2.2 AA accessibility, bilingual parity per the citizens chapter, clinical safety per DCB0129/0160, API-first per the target architecture. These are acceptance criteria, not aspirations — a service that fails them does not pass its assessment.
  7. Works in the open. Code open source by default at the shared layers; decision records published; every service has a public “what we’re building and why” page.

The consequence mechanism

Standards without spend controls are posters. The funding envelopes of Break the Annual Trap release money at assessment gates: no passed assessment, no next tranche. The assessment panel is drawn from outside the delivering body — health-board peers, the standards body, and at least one external assessor from a comparator jurisdiction. Assessment reports are published within 14 days, per Radical Transparency.

Why this is in the blueprint

Every mechanism above exists and works at scale — in GDS assessments, in NHS England’s service manual, in the DORA research corpus. None of it operates in NHS Wales today, and the record shows what its absence produces: 550 tracked milestones with no confidence scoring, programmes “live” in two boards of seven after six years, and delivery theatre standing in for delivery. The standard replaces the theatre with a script anyone can audit.