The single source of truth is the most popular phrase in enterprise data strategy and one of the least examined. Every data platform pitch deck contains it. Every data governance program claims to be building toward it. It's treated as self-evidently desirable — the goal state that justifies the investment.
The problem is that for most organizations, in most domains, it's a fiction. And building strategy around a fiction produces architectures that don't fit the problem.
Why Multiple Truths Exist
Consider a straightforward question: how many active customers does a Turkish insurance company have?
The actuarial team has an answer. The finance team has an answer. The regulatory reporting team has an answer. The CRM team has an answer. In all likelihood, these answers are different — and all of them are correct, given their respective definitions and purposes.
The actuarial answer reflects policies with active risk exposure. The finance answer reflects policies with outstanding premium receivables. The regulatory answer reflects the definition required by the relevant statutory framework. The CRM answer reflects customers who have engaged in the last twelve months.
These aren't errors. They're legitimate, purposeful interpretations of "active" that serve different business functions. Forcing them into a single definition doesn't resolve the tension — it just moves the argument about which definition wins into a governance committee where it will never be resolved.
What "Single Source of Truth" Actually Achieves
When data leaders talk about a single source of truth, what they usually mean — when they're being precise — is one of several things:
Authoritative source systems. There should be one system that is the master record for a given entity. Customer demographics live in the CRM. Policy data lives in the policy administration system. This is achievable and valuable.
Consistent definitions within a context. Within regulatory reporting, "active policy" should mean the same thing across all reports submitted to the same regulator. Within actuarial modeling, "exposure" should be defined consistently across models. Consistency within a context is achievable.
Agreed lineage. When a figure appears in an executive report, there should be a documented, agreed path from source to presentation. Nobody should be pulling the same KPI from two different places and getting different numbers.
These are all achievable and worth pursuing. They are not the same as claiming there is one definition of truth that serves all purposes simultaneously.
The Architecture That Follows From Honesty
If you accept that multiple legitimate truths exist, the architecture looks different than if you're pretending to build a single source.
You need semantic layers, not just data layers. The raw data is shared. The business definitions that transform raw data into business metrics are context-specific. A well-designed semantic layer makes this explicit: here is the data, here are the business rules that apply it to this context.
You need reconciliation infrastructure, not just integration. When two legitimate answers to the same question differ, you need to be able to explain why. That requires understanding both definitions, the data that feeds each, and the transformation logic that produces the difference. Without this infrastructure, every cross-functional discussion about the numbers becomes a fight about whose number is right.
You need explicit ownership of definitions, not just data. The actuarial team owns the actuarial definition of active customer. The regulatory team owns the regulatory definition. Both definitions are documented, versioned, and maintained. When either changes, the downstream calculations that depend on it change with it.
The Governance Implication
The single source of truth framing tends to produce governance structures that are trying to resolve a problem that doesn't need to be resolved. Committees that can't agree on a definition, because the definition is genuinely context-dependent. Endless debates about canonical metrics that produce no canonical answer.
The more productive framing: governance should define which team owns which definition, how definitions are documented, and what happens when definitions conflict in a specific report or decision. Not: what is the one true definition of this metric.
That's a harder conversation to have, because it requires admitting that complexity exists rather than promising to eliminate it. But it produces governance structures that actually govern.
The single source of truth is a useful aspiration for source system mastering and lineage clarity. It's a misleading aspiration for business definitions in complex organizations. The teams that build well are the ones who can hold both truths simultaneously: shared data, contextual meaning.