Measuring maturity

Everyone agrees data maturity matters. Nobody can measure it.

The argument has never been about importance. It is that the thing everyone agrees is important has never been measurable.

First, a definition

What data maturity actually means.

Data maturity is how far a data product can be depended on and proved — how well it is defined, owned and contracted, how reliably it runs, whether the governed steps actually happen, and whether there is standing evidence for all of it. Measured on the product, not on the programme.

It is a property of a product, not of a department

“The organisation is at level 2” is not actionable. “Credit Exposures Reporting scores 72, and here is the finding holding it back” is. The unit of measure is the data product, every time.

It is evidence, not opinion

Maturity is not what a team believes about itself. It is what can be shown: a named owner, a current contract, a quality result with a date, an approval that actually happened.

How we calculate it

Six steps, and nothing in them is a questionnaire.

The high-level shape of the calculation. Every step is deterministic — the same evidence produces the same number.

Step 1

Observe the evidence

The platform reads what it already holds — documentation, ownership, contracts, gate decisions, approvals — plus the job runs, quality results and asset events your own tools report in by API.

Step 2

Test it against capabilities

Evidence is evaluated against a canonical set of governance capabilities, anchored to established frameworks rather than invented for the occasion.

Step 3

Level each capability

Each capability gets a level from the evidence behind it. Where there is no observable evidence, it is reported unassessed — not level one, and not a failure.

Step 4

Roll up into four dimensions

Capability levels aggregate into product state, operations, audit and workflows — the four dimensions that carry the 0–100 score.

Step 5

Penalise the weakest, don’t average it away

The roll-up is deliberately not a plain mean. A single weak dimension pulls the score down rather than being offset by a strong one, because that is how the risk actually behaves.

Step 6

Re-compute on a cadence

Because nothing was asked of anyone, running it again costs nothing. The second run is what turns a number into a trend — and a trend is what a board can govern.

The model is not in this loop. The score, the findings and their severity are computed by rules. The AI advisor explains the result afterwards, citing the same evidence — it never produces the number.
The four ways organisations try today

And why each one fails the moment you need to act on the answer.

A consultant assessment

Six figures for a snapshot. Accurate the week it is written, stale by the time it is circulated, and impossible to repeat cheaply.

A self-assessment survey

Measures opinion, not evidence. Everyone rates themselves a three, and nobody can show why.

A framework scored by hand

DMBOK, DCAM and CMMI are sound. Scored manually they rate the programme, not the products — and “level 2” is not something anyone can act on.

A spreadsheet someone owns

New assessor, new answer. No evidence trail, no comparability, no trend.

We also sell an expert assessment. Here is how it answers this critique

The frameworks are not the problem. DMBOK, DCAM and CMMI are good frameworks — the problem is manual scoring at organisation level. All four methods above share the same three faults: they measure the organisation instead of the product, they capture opinion instead of evidence, and they produce a snapshot instead of a trend.
A number and a stamp

The score is computed from four dimensions the platform already observes.

Nobody fills in a questionnaire. The score reads what is already there.

72
Maturity score

0–100 per data product · illustrative

Product state

Documentation, ownership, contracts — is the product defined, owned and described well enough to be relied on?

Operations

Health, freshness, cost — is it actually running as promised, and at what price?

Audit

Evidence standing and current — is the proof still valid today, not just on the day it was filed?

Workflows

Gates passed, on time — did the governed steps actually happen, in order, when they were due?

The board question — “are our data products mature enough?” — becomes a number, per product, on a date. And because it is computed rather than asked, running it again next quarter costs nothing and produces a trend.
Certification

A stamp that is earned from evidence, not self-awarded.

Bronze to Gold are earned automatically from evidence the platform already holds. Platinum requires an independent audit.

Bronze

Documented and owned.

Earned from evidence

Silver

Monitored and contracted.

Earned from evidence

Gold

Evidence standing, gates enforced.

Earned from evidence

Platinum

Independently audited.

Independent audit required
That distinction matters to a regulated buyer: the stamp is not self-awarded. Every badge carries the criteria version it was assessed against and the date it was assessed, so it reads as auditable rather than decorative.

Get a score on your own data products.

A 30-day pilot, one slice of your estate, success criteria agreed up front — and you keep the result either way.