What breaks when a business definition changes
A definition is rarely stored in one place. When one copy moves and the others don't, the damage shows up far from the edit, and late.
Ask three teams in the same company what an “active customer” is and you will often get three answers. None of them is wrong in its own context. The problem starts when a number built on one answer is compared with a number built on another, and nobody knows they differ.
This post is about how that happens, and why it is so hard to see.
A definition lives in more places than you think
Take any metric that reaches a management report. Its definition usually exists in at least four forms:
- The requirement. A ticket or specification that says what should be counted and how.
- The implementation. A query, a pipeline step or a calculated field that does the counting.
- The explanation. A wiki page or policy document that tells people what the number means.
- The presentation. A dashboard or report label, sometimes with a footnote.
Each form is owned by a different person. Each is edited with a different tool. And none of them is aware of the others.
Increasingly there is a fifth form: the text an AI assistant finds when someone asks it a question. That copy has no owner at all.
How the copies drift apart
Drift usually starts with a sensible local decision.
An analyst tightens a filter to exclude test accounts. A product owner updates a ticket after a stakeholder conversation. A finance team changes the period a ratio is measured over to match a new reporting calendar. Each change is correct for the person making it.
Nobody tells the other people who depend on that definition.
Why the damage shows up late
The first change is rarely the expensive one. The person who made it understood it. The cost sits further downstream:
- Derived metrics. Anything calculated from the changed term now uses the new logic, whether its owners agreed or not.
- Work in progress. Tickets built against the old definition are now wrong, but they still look finished.
- Reports already issued. Last quarter’s figures used the old definition. This quarter’s use the new one. The comparison between them is no longer like for like.
None of these raises an error. The numbers still compute, but they now measure something different from what the reader assumes.
What it takes to catch it
Catching drift early needs three things that most organisations do not have in one place.
A reference version. Someone has to say which definition is the approved one. Without that, every copy is equally authoritative and “drift” has nothing to be measured against. This is the job of a business glossary with real ownership behind it.
A map of what depends on what. Knowing that a term changed is only useful if you also know what uses it. That is impact analysis, and it only works if the relationships are recorded before the change, not reconstructed after.
A way to put a number on it. “This change affects twelve things” rarely wins a prioritisation meeting. “This change will cost this much rework, and here is the arithmetic” usually does. We wrote about how that calculation works.
Where SynkBase fits
SynkBase is built around those three things. It keeps an approved version of each governed term, detects when a connected system moves away from it, traces what depends on it and prices the rework from your own delivery data.
We are building it with design partners. If the scenario above sounds familiar, we would like to hear about it.