A 40-person company has three different numbers for "completed order" living in three spreadsheets. Twelve people touch data at all. A consultant has proposed a governance programme. Which arrangement fixes the problem with the least machinery?

The deciding property

There is exactly one disagreement, and twelve people who can all be in the same conversation. At that size the problem is not coordination cost, discovery or audit — it is that three people each wrote their own SQL and nobody decided which one is right. The fix has to do two things: settle the definition, and make the settled version the easiest one to use.

Why one owner and one view

A shared view is an enforcement point: the definition exists as the code everyone queries, so using the agreed number is less work than rewriting it. That inverts the usual governance economics, where the compliant path is slower than the shortcut.

A named owner gives the definition somewhere to be changed. When finance wants cancelled-within-24-hours excluded, there is one person who decides and one object to edit, and the change reaches all three spreadsheets at once because they are now reading the same view.

The whole thing costs an afternoon. Compare that with a council: 26 meetings a year with six people is roughly 156 person-hours before anything is written down.

Why the other options fail

• The stewardship council. It is what you build when definitions are contested across business units with different incentives and someone needs standing to rule. With twelve data users and one disputed term, the council spends its first meeting on a terms of reference document, and the three spreadsheets are unchanged a quarter later. A council produces decisions, not implementations.
• The catalogue with a harvester. Harvesting is the right call at thousands of tables, where nobody can know what exists. Here everyone already knows where the data is. The catalogue will faithfully document all three conflicting definitions and resolve none of them, which makes the disagreement look official.
• The written policy. A policy says what should be true. Nothing checks it, so the three spreadsheets stay as they are and the organisation now believes it has addressed the issue. This is the most common failure in small-company governance: an artefact whose existence substitutes for the fix.
• The two-approver certification workflow. Approval gates on report sharing buy auditability and cost every analyst days of waiting. With twelve users the queue is the whole bottleneck, people route around it by sending screenshots, and you have lost visibility rather than gained control.

What it costs, and the failure mode

Someone will copy the view's SQL into a new spreadsheet, usually to add a filter. You will not notice from the view. The signal is the query log: if a table the view reads from is being queried directly by people who should be using the view, a second definition is forming. Checking that once a quarter takes minutes.

When this is the wrong answer

Three conditions flip it. Above roughly 30 to 60 governed definitions, one owner cannot hold them and you need distributed ownership with an escalation path. When two business units genuinely need different definitions — finance's revenue and product's revenue are both correct — you need a glossary that records both with their scope rather than a single view that pretends one wins. And in regulated reporting, the approval record is itself the deliverable, so the certification workflow is not overhead, it is the product.