How the Transformation Index is built

This page is for anyone who would rather check the reasoning than take a score on trust. It sets out what the Index measures, how it is weighted, what it asks you at the start and why, and where the frameworks you already know genuinely fit against it. It also says plainly where they do not.

Nineteen dimensions, seven domains

The instrument scores nineteen dimensions. They are grouped into seven domains for reporting, so an executive summary reads at domain level and the detail sits one click below it. The grouping is presentation. The scoring happens at dimension level and never at domain level, because that is where the evidence actually attaches.

Every dimension carries the same weight in the composite. One nineteenth each. That is a deliberate choice rather than an unfinished one: a weighting that says data matters more than governance is a claim about your organisation, not about the instrument, and it is the kind of claim that quietly decides the answer before anybody has been asked a question. A domain's arc on the wheel is therefore sized by how many dimensions sit inside it, and nothing else.

If a dimension does not apply to you, it is marked not applicable and removed from the composite rather than scored zero. A zero is a finding. A blank is not, and the two should never be averaged together.

The bottleneck rule

A composite flatters. An organisation can average 58 while the one capability everything else waits on sits at 31, and the average will not say so. So readiness to scale is reported as the minimum of the critical dimensions, never their mean.

A composite of 72 with data at 20 is 20-ready. That single rule changes more investment decisions than any other part of the method.

This is also why the report names one constraint rather than handing back a ranked list of nineteen. A list invites you to work on the three easiest. The rule points at the one that is capping the rest.

What intake asks, and what each answer changes

Four things are asked before the questions start. Each one changes something specific in the output. Nothing is collected because it would be interesting to know.

What is deliberately not asked. Geography, until the external benchmarks are split by region. Personal seniority or tenure, because the persona already carries the real authority signal and an individual's tenure currently feeds nothing. A field with nothing behind it is a field that makes a profile look thorough and changes no output, and there are enough of those in this category already.

Where Kotter, McKinsey 7S and Porter fit

Three well known frameworks, three different verdicts. Presented separately rather than force-fitted, because two of them fit and one of them only half does.

Kotter’s eight steps: a good fit

Every one of the eight steps can be answered from evidence the diagnostic already gathers, just organised differently. Urgency comes from strategy and value. The guiding coalition comes from the real stakeholder map. Removing barriers comes from the open recommendations. Short term wins come from whatever the roadmap actually lands inside thirty days. Institutionalising change comes from the change and adoption maturity ladder itself. Read this way it becomes an eight row checklist where every row is backed by a real answer, rather than a motivational sequence.

McKinsey 7S: five of seven, honestly

Said plainly: the people dimension already combines Staff and Skills into one score, so separating them here would be presentation rather than measurement. And nothing in nineteen dimensions measures leadership style. A 7S view built on this instrument should show five real bars and mark Style as not assessed, rather than invent a seventh score to complete the picture. The same rule applies everywhere else here: where there is no credible measure, the gap is shown rather than filled.

Porter: the Value Chain fits, the Five Forces does not

Five Forces describes the market outside the organisation. Rivalry, supplier power, buyer power, substitutes and new entrants cannot honestly be produced from nineteen internal capability scores, and any tool that claims to is generating them. It belongs at discovery, as a handful of context questions you answer directly, and then as narrative context against the strategic choices. Never as a computed score.

The Value Chain does fit, because it is internal. Tagging each use case with the activity it touches lets the portfolio group by which part of the business it actually helps, rather than presenting as a flat list.

The product takes that further. The seven gates are themselves a value chain, and the dashboard reads the same nineteen dimension scores through them: each gate scores as the weighted mean of the dimensions inside it, on exactly the weighting the scorecard uses, and the lowest scoring gate caps every gate downstream of it. That is the bottleneck rule again, applied along a sequence rather than across a set. The mapping is fixed and every dimension sits in exactly one gate, so the two lenses cannot disagree about a score. Two placements are a judgement call rather than obvious and are named as such on the pane itself: operating model sits at Deployment because it decides whether work lands and runs, and the product operating model sits at Dividend because it decides whether anyone owns the thing after go-live.

The Value Ceiling, stated as arithmetic

For anyone who would rather check the working than accept the figure, the whole calculation is four lines.

1. Gate score is the weighted mean of the dimensions inside it, on exactly the weighting the composite uses. Nothing is reweighted for this view.
2. The ceiling is the lowest gate score. This is the bottleneck rule already described above, applied along the sequence instead of across the set.
3. What is sitting idle is the strength built in steps past your weak link, as a share of everything measured, applied to your committed figure. No realisation rate is assumed anywhere, because there is no evidence for one.
4. The next constraint is whichever gate sits at the next distinct level up. No simulation and no judgement enters here: sorting the gates by score produces the sequence.

Two known biases in this instrument, measured and published

Both were found by testing the instrument against itself rather than by a client complaining, and neither is fixed by wording, so they are stated here instead.

Gates built from fewer dimensions are named the constraint more often. A gate score is the mean of the dimensions inside it, and a mean of one number varies more than a mean of four. Run twenty thousand random profiles through the mapping and the single dimension gate, Data, comes out lowest 26.6 per cent of the time against 7.5 per cent for the four dimension gates, where an unbiased instrument would give 14.3 per cent each. That spread is arithmetic, not evidence. It does not make a Data finding wrong, but it does mean Data starts with an advantage, and a close call between a small gate and a large one should be read as a tie. The report now says so on the face of it whenever the gap between the two lowest gates is five points or less.

No single person can answer for all nineteen dimensions. The interview only asks what a persona can credibly speak to, so a CIO is asked about fourteen dimensions and an HR director about five. Earlier versions filled the rest with an estimate and let those estimates into the gate arithmetic, which meant the named constraint could rest partly on numbers nobody had given. That is now fixed: a dimension nobody answered is excluded rather than estimated, a gate with nothing behind it is marked not assessed and can never be named as your constraint, and the report states how many of the seven gates actually had answers behind them. The honest consequence is that a single respondent assessment covers four to seven gates, not seven, and the constraint may sit behind a gate that person could not speak to. More respondents from the same organisation is the fix, and the product tells you when you need them rather than papering over it.

Every step here is arithmetic on your own answers and your own committed figure. The judgement calls are the gate mapping and the order of the seven steps, both of which are published above so you can disagree with them specifically rather than in general.

If you did not give a committed figure at intake, none of this is invented on your behalf. The ceiling is reported in points, the value pool is labelled illustrative, and the report says so.

Discovery through to value

1. Discover. Industry, the problem in a sentence, persona, and the four intake answers above. This is what makes the interview adaptive rather than a fixed questionnaire.
2. Diagnose. The adaptive interview, evidence tier recorded at the point each answer is given, contradictions surfaced rather than averaged away.
3. Design. Target operating model against the maturity ladders, root cause traced to one validated constraint, risks with owners and guardrails.
4. Decide. Strategic choices the evidence actually forces, decision rights written down, and what is explicitly being deferred.
5. Deliver and learn. A sequenced roadmap, value tracked through its proof stages rather than as one number, and a next cycle that can be compared against this one.

Every stage names what it cannot tell you. That is not modesty. It is the only honest way to read a diagnostic where most of the input is what somebody told you.