How this scan measures

What this scan measures

The Data Maturity Scan measures along four dimensions, with four questions per dimension:

• Direction and ownership: where do decisions about data come from, and is anyone identifiably accountable?
• Reliability of figures: do the figures carry what is built on them?
• Availability and foundation: is the data there when it is needed?
• Use in decision-making: is anything actually decided with it?

The questions are deliberately about observable behaviour rather than self-image. Not “how mature do you consider yourself”, but: what happens when two departments quote a different number? How old is the information your most recurring decision rests on? Who can get an answer out of the data without help?

How scoring works

Per dimension the average is taken of the questions answered. That average falls into one of five levels: Ad hoc, Reactive, Structured, Proactive, Data-driven.

You get no overall score. That is a deliberate choice, in line with the UK government method for data maturity and with Microsoft, both of which prescribe that levels are determined per component and not added up. An average across four dimensions hides exactly what you want to know: where things are out of balance.

In case of doubt a level rounds down. You reach a level only when three of the four answers sit there and the fourth is at most one step lower. That is strict, and deliberately so: a level you half reach is not a level.

Alongside the level profile you get a pattern, the shape your answers form together, and, where there is reason for it, one or two priorities. If everything sits close together, or if you score high across the board, no priority is named. Inventing a priority where there is none makes the advice worse.

“I don’t know” is a valid answer

On every question you can indicate that you do not know the answer. It does not count as a zero and it does not push your result down; the question simply falls outside the calculation. If you do not know more than one answer within a single dimension, that dimension gets no level and the scan says so.

That is not a shortcoming but a finding in itself: it means this information is not within reach of someone in your role. In many organisations that is precisely the problem.

AI is a conclusion, not a dimension

An earlier version of this scan counted AI as an axis on which you could earn points. The result: an organisation deliberately not doing AI was penalised for it. That is wrong.

AI readiness now follows from the foundation and the reliability of the figures. That matches the NIST AI Risk Management Framework, which requires the non-AI alternative to be weighed in rather than left out.

What this scan is not

This is a self-assessment, completed by one person. Self-assessment and actual performance demonstrably diverge on complex subjects. The outcome is enough for a direction and a conversation, not for a judgement or an audit.

We show no comparison with other organisations. We have too few completed scans for that, and a benchmark on a handful of entries is misleading. As soon as that changes, we will publish the numbers alongside.

The scan does not replace an assessment in which several people from your organisation answer independently. It is precisely the difference between those answers that often produces the best conversation.

What we store

Your answers are stored anonymously, with the version number of the instrument and of the texts. Without a name, without an email address, without an IP address. We use them to improve the scan: which question is often skipped, which outcomes occur.

If you request the report, you supply your details yourself. That flow is separate from the stored answers and is not linked to them.

Sources

The design is informed by the UK government maturity model (Data Maturity Assessment), the CMMI Data Management Maturity model, the adoption roadmap for Microsoft Fabric and the NIST AI Risk Management Framework. The scan is not affiliated with any of them as a certification or formal assessment.

---

Version 2.0.0 · revised on 26 August 2026. Every revision of the questions or the scoring changes this version number, so that outcomes from different periods can be told apart.

← Back to the scan