Radical Transparency

Live delivery dashboard showing programme status in real time. Outcome-only KPIs replacing activity metrics. Published programme costs alongside adoption metrics — the public should see: “OpenEyes: £8.5M spent, live in 2 of 7 health boards after 6 years.”

Published vendor contract values and delivery performance. No more unnamed suppliers on £47-56M contracts. Staff-reported delivery confidence published alongside management-reported programme status — if there’s a gap, the staff number is more reliable.

That is the headline. The operational specification below makes it stick.

What gets published, when

The diagnosis identifies seven categories DHCW has actively suppressed. Each requires a named publication cadence:

1. Whistleblowing data — monthly count by category and outcome status; 30-day maximum publication lag; year-on-year trend visible.
2. Disciplinary proceedings — monthly count by category and band, anonymised; outcomes tracked separately from initiations, so pretextual use becomes visible as a pattern.
3. Staff leavers — monthly leaver analysis by reason, role, and destination; 12-month rolling trend. The L9 finding that “the absence of data is itself data” closes by default. Together with the burnout, sickness, and PADR figures published alongside, this makes the structural inverse of psychological safety — currently sustained by stripping every relevant figure from the published record — impossible to maintain.
4. Contract values — every contract above £100K published within 30 days of signature; sole-bidder contracts above £100K published within 14 days with a public justification. The RISP (£47-56M), Channel 3/Aire Logic, and Promptly Health (£11M, no published business case) undisclosed-value gap closes by default.
5. Programme delivery status — live dashboard, milestone-level, weekly update; current state, slippage against the previous baseline, named accountable individual.
6. Confidence-of-delivery scores — every milestone owner attaches a confidence score (1-5) at the start of each reporting period; both the score and the ground-truth outcome are published, so optimism bias becomes measurable. This is the direct counter to Evans’s admission that “confidence of delivery is not currently measured” across 550 tracked milestones.
7. Board and committee minutes — full publication within 14 days of meeting; original transcript form, with personal information redacted only as required by data protection law; no editorial sanitisation between transcript and published record.

Statutory backing

Each of the seven categories above is published under a statutory duty, not a discretionary policy. This is the highest-leverage element of the intervention: removing leadership discretion converts publication from a governance negotiation into a legal compliance matter. Non-publication becomes a matter for the courts and the Information Commissioner, not for internal escalation. This single design choice neutralises the most reliable strategy of L10: The Information Fortress — it removes the option of withholding.

Anti-sanitisation protocol

The diagnosis documents 107 sanitisation instances, 237 hiding-intent passages, and a curation ratio that fell to 10.7% in one published transcript — 3,680 published words from 34,257 spoken, with sixteen of nineteen speakers removed entirely. The transparency intervention must directly counter the curation mechanism, not merely publish more data on top of it. Three rules:

• Minimum curation ratio of 60% across published meeting minutes. Below that threshold, an explanation accompanies the published version naming who curated and why.
• Named-individual challenges preserved verbatim. Where a board or committee member raised a question or objection, it is published as spoken, not paraphrased. Chair Jones’s “heap misery on misery” warning, Lloyd-Jones’s “anodyne phrase” candour about technical debt, Evans’s “biggest disappointment” admission about milestone owners lying — none of these would have been deletable under this protocol.
• Audit trail of edits. Every change between transcript and published minute is logged with editor, timestamp, and reason. Audit Wales receives the full audit trail quarterly.

FOI and the cost of refusal

Every refused FOI request is published with the reason for refusal. ICO referrals are tracked publicly. Cabinet Office FOI compliance scores are published quarterly. Refusal itself becomes public data, which makes refusal a reputational cost rather than a costless default. This addresses the documented pattern of FOI requests being declined or delayed without consequence.

Dashboard design integrity

The dashboard is itself a target for capture. Three rules prevent it:

• Published technical schemas. What is on the dashboard, where the data comes from, how it is computed — all open. External developers can replicate the dashboard from raw data and verify the result matches.
• External audit verification. A named external body — Audit Wales or its commissioned digital expertise — verifies the dashboard against the underlying data on a quarterly cycle.
• No “context” columns. If “context” or “narrative” columns are added to soften or qualify the headline numbers after launch, the dashboard has been captured. The trigger condition for governance escalation is automatic.

Why this intervention without I1 fails

A transparency dashboard launched under leadership that blocked carenhs.org on the NHS Wales network and published zero whistleblowing statistics for five years will be designed to satisfy the letter of the rule and not the spirit. Implementation under the existing leadership is not an option — it is a guaranteed capture point. This is why Competent Leadership is the prerequisite. The dashboard must be designed and operated by leadership accountable for what it shows.

This intervention dismantles L6: The Manufactured Narrative and L10: The Information Fortress by removing the discretion the manufacture and the fortress depend on.