Your transformation may be reporting progress. But can the evidence support the confidence?

Transformation Recovery OS is an evidence-led decision and recovery methodology for complex programmes where conventional status reporting is not enough to determine whether the intended outcome is genuinely secure.

The problem

In practice, that means asking three different questions of a programme, not one.

How it thinks

I follow the same five moves on every engagement:

Evidence → Pattern → Causality → Recovery

Decision states protect the business outcome.

Every assessment resolves to one of six states:

What makes it different

Choose your route

What is already putting your transformation at risk?

A short check for early signals across delivery, capability, readiness and dependencies.

Not knowing whether a material control or condition exists can itself be useful information. Select “Unsure” where you cannot confidently assess an area.

Choose the diagnostic that matches your context.

Both routes use the same evidence-led logic, but the questions and outputs are tailored to the decision-maker.

The diagnostic is designed to assess programme conditions without requiring commercially sensitive programme documents during the self-service assessment.

Separate reported progress from evidence-supported readiness.

A worked example. No client or programme is identified.

The figures below come from a worked example, chosen to show the shape of the output. In a real report every number traces back to the specific answers given in the assessment, and the report shows that working.

Move from diagnosis to focused intervention.

Playbooks are prescribed from the Recovery Report rather than presented as generic downloads.

Most methodologies ask you to trust them. This one shows its working.

Transformation Recovery OS separates reported progress from evidence-supported readiness. That is a claim about judgement, and a claim about judgement should be tested rather than asserted.

So it was tested against major UK government transformation programmes across eleven sectors: each assessed at a fixed point before its outcome was known, with the expected answer written down and locked before the engine was run. This page sets out how that was done, what it found, and where the instrument stops and experienced judgement takes over.

1. Why this page exists

Every transformation methodology on the market describes itself as rigorous, evidence-led and proven. Almost none of them show their working.

That is a problem for a methodology whose central principle is that confidence is not evidence. If Recovery OS cannot survive the standard it applies to programmes, it has no business applying that standard to yours.

2. How it was tested

A set of major UK government transformation programmes was assessed under a preregistered protocol. Each was reconstructed at a fixed point in its life, at a moment where a real decision had to be made, and every assessment answered the same question: given only what could genuinely have been demonstrated at that moment, what was the right call?

The protocol has four rules, and they are what make the result meaningful rather than decorative:

They span eleven sectors: transport, identity, health, agriculture, pensions, energy, emergency services, public health, rail, defence and immigration.

Each was benchmarked against the independent public record of what actually happened, drawn from National Audit Office and Public Accounts Committee reporting published after the assessment date.

Why government programmes

Public sector programmes are the hardest available test, not the easiest. They are the only large transformations whose failures are independently investigated and published in forensic detail, by auditors with statutory access to the programme’s own records. A private-sector programme of the same size and cost is rarely examined at all, and almost never in public. Testing against the National Audit Office record means testing against a benchmark nobody involved could edit after the fact.

The instrument itself is sector-neutral. Its thirty-nine items ask about evidence, dependencies, capability, testing, governance and adoption. None of them asks what industry the programme is in, and none of the scoring changes by sector.

Financial services is the next addition to the tested set, and the material exists to do it properly: regulatory final notices and commissioned independent reviews provide the same quality of post-hoc record that the National Audit Office provides for government. That work is planned, and its results will be published here in the same form as the results below.

3. What the testing found

The figure that matters is the second. Several of the programmes assessed went on to fail publicly and expensively. In not one case did the methodology conclude that a troubled programme was healthy.

That is the error that costs money, and it did not occur.

Where the assessment differed from the independent record, it differed the same way every time: it was more cautious than reviewers who had the benefit of hindsight. That direction is a design position, not an accident. An instrument advising on irreversible decisions should be harder to satisfy than the decision itself.

Its recommendations were consistently actionable. In every case the first recommended action named an accountable owner and a conditional path forward, and in almost all of them the specific evidence that would change the conclusion. Those are the properties that make a recommendation act on, rather than merely agree with.

On the headline decision it either agreed with the independent record or was more conservative, in every case, without exception. Section 4 sets out where that leaves the boundary between the instrument and the judgement of the person using it.

4. What it settles, and where your judgement takes over

5. An earlier result, held to the same standard

An earlier assessment examined TSB’s 2018 core banking migration at a fixed point twelve days before go-live, using only facts establishable at that date. The assessor did not know how it ended. Both the practitioner and the engine independently reached Pause. The migration went ahead; within hours customers were locked out, and the cost was approximately £366 million.

It is excluded from every figure on this page, for one reason: the individual responses behind it were never saved. Without them the assessment cannot be re-run, re-checked, or tested against a corrected version of the methodology.

A conclusion whose working no longer exists is precisely what this methodology exists to challenge. It is held to the same standard here, and retained only as an archived reference.

6. What the testing changed

Two earlier versions of the methodology were suspended as a result of their own testing. Version 1.0 was withdrawn after an assessment of Universal Credit exposed a defect in how it weighted causes by delivery stage. Version 1.1 was withdrawn after a deliberate positive control (running it against the London 2012 delivery programme, which succeeded) surfaced an over-caution defect before any client had seen the methodology. That failure was traced to a specific rule, corrected, and the correction tested against controlled probes before the version was frozen.

The programme of held-out cases then measured precisely how far that caution extends, published in section 3. That measurement is what drives the next refinement.

The standard applied is that a change to the methodology requires the same problem to appear twice, the full regression suite to pass, and every prior case to be re-run and explained. One proposed improvement was built, tested, and reverted when it altered an earlier diagnosis in a way that could not be justified.

7. The honest claim