Monitoring Framework

Four quadrants, tracked monthly:

Quarterly assessment: Has Cluster B been broken? Which Cluster A loop is currently strongest? Has L2: Credibility Spiral reversed? Are new loops emerging?

Red flags: Dashboard launch delayed past month 9. “Context” columns added to soften delivery numbers. Embedded team leads hired from within DHCW. Multi-year funding allocated to the organisation rather than to programmes. Curation ratio of published board minutes drops below 60%. Any new “declared nil” entry against a director with a known undisclosed interest. “Receive and note” cadence above 50% of substantive board agenda items. Audit Wales positive assessment within six months of a worsening of measurable indicators (the captured-assurance pattern from 2025). Approval-without-scrutiny on any contract above £5M. Recruitment freezes paired with milestone tightening (the 2026 remit-letter pattern). Staff-welfare suppression — burnout figures absent from published staff-survey items in board minutes; year-on-year burnout change stripped; sickness trajectory not published alongside headcount; any board action item recorded as “Received & Discussed. Action: None to note.” against a worsening staff-welfare indicator. Infrastructure-incident recurrence — any second occurrence of the same infrastructure failure pattern within 18 months without a published corrective-action close-out (the 2024-2025 data centre cooling failover pattern). Any of these signals that the reform has been captured by the system it was designed to change.

Reform-the-Funder leading indicators (Intervention 6): RAG-rating audit trail visible to the Senedd Public Accounts Committee monthly, with every rating change logged by author, date, and rationale. Capital-allocation timing relative to programme inception (capital settled before fiscal year start, not later). Remit-letter content scored against milestone load — any reduction in input authority paired with proportionate output relief. Frequency of Welsh Government named in published programme post-mortems alongside DHCW.

When an intervention appears not to be working, use this decision tree before concluding it has failed:

1. Has the intervention been in place for less than its expected delay period? (Trust rebuilds on a 2-5 year lag; embedded teams need 6-12 months to ship.) If yes — check leading indicators. If leading indicators are moving in the right direction, the intervention is working. Be patient.
2. Are leading indicators flat or declining? Check whether Cluster B is blocking — is information being suppressed about the intervention’s progress? Is the intervention being undermined from within? If yes — the problem is not the intervention. It is incomplete implementation of Intervention 1.
3. Are leading indicators flat with no evidence of Cluster B blocking? The intervention may need adjustment. Commission an independent review of that specific intervention, not a wholesale rethink.

The most likely failure mode is premature abandonment: people judge the reform too early, see no results, and abandon it. Every delay in this system is measured in years, not quarters. The decision tree pre-empts this.

The critical signal: One health board saying “DHCW delivered what they promised.” That is the single most important indicator. Everything else is leading. That is the outcome.