Microsoft Fabric IQ can model how your business works. It cannot agree what your numbers mean. Here is the five-minute test to run before you license it.

Microsoft Fabric IQ, and the dictionary nobody wants to write

Your database does not know what your business means by an active customer. Neither does the next platform you buy. That knowledge lives in people's heads, in three or four slightly different versions, and it has never been written down. Microsoft Fabric IQ is the latest and best attempt to fix that: an ontology layer that holds the meaning of your business in one place, so that people and AI agents can both reason in your language rather than in table names.

This connects with our Microsoft Fabric delivery and, separately, with The Data Engineering Layer That Decides Everything.

It is a good idea, and we have argued for the principle for years. But it walks straight into the same wall every tool of its kind hits. The software is the easy part. Agreeing what goes into it is the work, and most organisations have never done it.

A five-minute test that will unsettle you

Before you license anything, try this. Get your senior team to write down, separately and without conferring, their definition of three things: an active customer, revenue, and a completed project.

If the answers match, you are ready for Fabric IQ and most of the value is already yours. In fifteen years we have not once seen them match. That gap is the project, and no ontology layer can close it for you, because it is not a data problem. It is a business that has never been made to decide what it means.

The fights are always in the edge cases

Nobody argues about the headline. Everyone agrees revenue is money the business earned. The disagreement hides in the detail, and the detail is where a definition holds or falls apart. Does a customer who churned and came back count as new or returning. Is revenue booked at order, at delivery, or at invoice. Is a project finished at handover or at final payment.

We lost an afternoon with a client to that last one. Three people, three answers, each one already wired into a different report. None of them wrong, just never reconciled. Until that afternoon, every figure the business published rested on an assumption nobody had agreed to.

You do not settle this by committee, because committees produce a fudge that pleases nobody and means less than either original. You settle it by giving each definition a single named owner who decides, writes down the awkward cases, and is allowed to be unpopular. That ownership is the quiet machinery underneath any semantic layer that actually works.

What it costs to skip it

Skip the dictionary and nothing breaks loudly. There is no outage. The business simply stops trusting its own numbers, one meeting at a time. A year on, the board pack and the sales report disagree on the same month, nobody can say which is right, and every discussion carries a small tax of people arguing about whose figure to believe instead of what to do. That tax compounds, and the invoice never arrives in a form you can see.

Run the three-word test this week. Whatever it tells you will be the most useful thing you learn about your data this quarter, and it costs nothing but the nerve to ask.

When you are ready to build on it, that groundwork is where our Microsoft Fabric work and our analytics strategy engagements both start: the dictionary first, the platform second. Fabric IQ earns its name once the meaning exists. Before that, it is just a tidier place to be wrong.

If any of this sounds familiar, talk to us about your data.

Related reading

• Are your people happy? The first question in any Data or AI project
• Query Where Your Data Lives
• The Spreadsheet That Became A System Nobody Maintains