Why your business has four hundred versions of the same report

Self-service analytics solved a real problem. Business teams stopped waiting months for a report. What arrived instead was a tenant with hundreds of reports, several answers to every question, and no way to tell which one to use.

Why people rebuild rather than reuse

Not laziness, and not ignorance of what exists. Four reasons, in rough order of frequency:

They cannot tell if the existing one is right. No owner, no documentation, no indication of whether the logic matches the definition they need. Rebuilding is faster than investigating.

It is nearly right. The report exists but filters differently, or excludes something, or uses last year’s organizational structure. Copying and adjusting takes minutes.

They cannot get access. Requesting access has friction, and extracting the data and building their own does not.

Their definition genuinely differs. Sales counts bookings, finance counts recognized revenue. Both are correct. Neither is documented, so the disagreement surfaces as two reports rather than as a definition.

The underlying cause is modelling, not behaviour

Where every report connects to source systems on its own, every report implements its own version of every calculation. Two hundred reports means two hundred implementations of revenue, and they will not all agree.

This is not a discipline failure. It is the predictable result of a structure where there is nothing shared to build on.

What a governed model changes

One semantic model, built once, where the calculations live. Reports connect to it rather than to the raw sources. The definition of revenue exists in one place, and when it changes, it changes everywhere.

Self-service continues — people still build their own views, slice their own way, answer their own questions. They just do it on a foundation that is already correct, which is both faster for them and consistent for everyone else.

The definitions have to be settled first

The technical work is the smaller half. The larger half is getting sales, finance and operations to agree on what a customer is, when revenue counts, and which organizational hierarchy applies.

These conversations are slow and they are the actual deliverable. A model built on unagreed definitions just moves the argument.

Where a genuine difference exists — recognized revenue and bookings are not the same thing — the answer is two named measures, both defined, rather than one contested one.

Certification only works if someone owns it

Marking a dataset as certified means nothing unless there is an owner who keeps it current and a process behind the label. An estate full of certified content nobody maintains is worse than none, because it teaches people the label is meaningless.

Start with a small number of certified models that are genuinely maintained, and expand from there.

The test for whether it worked

Ask three people in different functions for last quarter’s revenue. If the numbers match, or the differences are explainable in one sentence by reference to a written definition, the model is doing its job.