Data governance & quality

The problem this solves

The symptom is familiar: two teams present different figures for the same measure in the same meeting, and the next twenty minutes go on reconciliation instead of the decision. Or a dashboard estate has grown to the point where nobody knows which version is current, so people quietly go back to their own spreadsheet.

Underneath it is almost always the same thing. Definitions were never agreed, ownership was never assigned, and nothing was written down. That is fixable, and it does not require a governance framework that takes a year to roll out.

The trap on the other side is over-correction: a heavyweight programme with councils, committees and a fifty-page policy that nobody reads. The useful version is proportionate. A fifty-person business needs a metric dictionary and a named owner per domain. A global function needs considerably more. Both need something people will actually use.

What I do

Reporting and data audit

A full inventory of what is currently produced: reports, dashboards, recurring extracts and the spreadsheets that have quietly become critical. For each, who uses it, what it claims to measure, where the data comes from and whether it is still needed. This alone usually finds a surprising amount that can be retired.

Metric definitions

The core of the work. Getting the people who disagree about a measure into a room and producing one agreed, written definition: the calculation, the filters, the exclusions, the grain and the owner. Where genuinely different definitions are needed for different purposes, they get different names, which removes most of the confusion at a stroke.

Lineage and documentation

A traceable path from source system to the number on screen, documented somewhere people can find it. Usually Confluence or your existing knowledge base rather than a new tool, because a documentation system nobody visits is worse than none.

Data quality assessment

Profiling the data behind your key measures for completeness, consistency, duplication and timeliness. The output is a prioritised list of quality issues ranked by the damage they do, so effort goes where it matters rather than everywhere at once.

Ownership and stewardship

Naming who owns which data domain, what that responsibility actually involves week to week, and how changes get proposed and approved. Kept deliberately light so it survives contact with people's day jobs.

Reporting standards

Conventions for naming, certification, publishing, versioning and retirement, plus a design standard so reports look consistent. Enough structure that the estate stays navigable, not so much that building anything becomes a chore.

How an engagement runs

What you get

• A reporting inventory with recommendations on what to keep, merge or retire
• A metric dictionary of agreed definitions, owners and calculations
• Lineage documentation from source to report for key measures
• A prioritised data quality register ranked by business impact
• An ownership model naming who is responsible for what
• Reporting standards covering naming, certification and publishing

Questions I get asked

Is this not just a documentation exercise?

Documentation is the artefact, but the value is in the agreements it records. The hard part is getting finance and operations to settle on one definition of an active customer. Once that conversation has happened, writing it down is the easy bit.

We are too small for governance.

Small organisations need less of it, not none. For a small team this can be a single well-maintained page of definitions and a named owner. The cost of skipping it entirely tends to arrive later, when reporting has multiplied and nobody remembers why two numbers differ.

Will this slow everyone down?

Badly-designed governance does. The test I use is whether a rule prevents a real problem that has actually occurred. If it does not, it does not go in.

Do we need to buy a governance tool?

Usually not to begin with. Most organisations get further with a well-maintained dictionary in the tools they already have. Buy a catalogue when you have outgrown that, not before.