When three analysts pull the same data and get three different answers, you don't have a data problem. You have a decision problem about who owns the definition.
Every leadership team we have worked with has lived through the same meeting. A senior executive asks a question. Three different analysts answer it with three slightly different numbers. Each number is defensible. None of them match. The room spends 20 minutes deciding which one to trust, and the actual decision the meeting was supposed to make gets pushed to next week.
Most companies treat that recurring meeting as a data quality problem. It is not. It is a decision-rights problem about who owns the definition of the metric. Until that ownership is settled, no amount of warehouse engineering, dashboard polish, or AI tooling will produce a consistent answer — because the underlying definitions are inconsistent on purpose.
What 'multiple versions of the truth' actually costs
The visible cost is the meeting-time tax: 15-20 minutes of every leadership meeting spent reconciling numbers instead of making decisions. Across a mid-market leadership team meeting twice a week, that is roughly 30-40 hours per quarter of the most expensive people in the company doing reconciliation rather than running the business.
The invisible cost is bigger. When numbers can't be trusted, decisions get deferred. When decisions get deferred, organizations slow down. When organizations slow down, the company misses windows that competitors don't. We have walked into engagements where the data quality complaint was the surface symptom of a six-month strategic decision that nobody could close because no one trusted the inputs. The cost of that delay almost always dwarfed the cost of fixing the underlying definition problem.
Why three numbers exist in the first place
When we audit the source of the disagreement, the three numbers almost always come from three reasonable but different choices. Finance counts revenue when invoiced. Sales counts revenue when the deal closes. Customer success counts revenue when the service goes live. All three are correct under their own definition. None of them are wrong. And none of them produce the same number for any given period.
Common definitional fractures
- Revenue: invoiced vs. booked vs. recognized vs. realized
- Pipeline: every stage-2 vs. opportunity-weighted vs. forecast-committed
- Active customer: any open contract vs. paid in last 90 days vs. logged in this month
- Churn: account-level vs. logo-level vs. revenue-weighted
- Headcount: payroll vs. badge access vs. contractor-inclusive
- Margin: gross vs. contribution vs. fully-loaded with allocated overhead
The decision-rights fix
The fix is not a better warehouse. The fix is a decision: for each headline metric, one named executive owns the canonical definition. The owner is empowered to decide which of the three reasonable interpretations is the company's official answer. Everyone else either uses that answer or labels their alternative explicitly as a non-canonical view.
A worked example
On a recent engagement, three teams were producing three different new-business numbers. Finance said new business was $4.1M for the quarter. Sales said $4.7M. Customer success said $3.9M. Three rooms, three numbers, three different definitions of when a customer 'started.' The CFO had been working around the disagreement for two years by triangulating between the three.
The fix took 90 minutes. The CEO named the CFO as the owner of new-business revenue. The CFO chose the recognized-revenue definition as canonical. Sales and customer success retained their own views but agreed to label them explicitly as their internal operating metrics. Within a quarter, leadership meetings stopped opening with the reconciliation conversation. The data team did not write a single new query.
Why the fix gets resisted
Naming a single owner for a metric forces a political conversation that everyone has been quietly avoiding. The team whose definition does not become canonical loses some authority over their function's narrative. The executive who takes ownership becomes accountable for defending the number in front of the board, which they might prefer not to be on the hook for. The data team, paradoxically, often resists because the multi-definition status quo creates demand for their reconciliation work.
The CEO is usually the only person who can force the decision. We have rarely seen this work without the CEO being explicit that the company is going to pick a definition and live with the consequences. The middle of the organization will optimize around the ambiguity if it is allowed to.
Anti-patterns we cut on every engagement
- Trying to make all three definitions produce the same number — they shouldn't, they answer different questions
- Hiring a data engineer to 'reconcile' the differences — the differences are real, not data errors
- Buying a new BI tool to 'unify the truth' — the tool cannot adjudicate decision rights
- Forming a metrics committee with no executive on it — the committee cannot make the call
- Letting each function keep their definition without labeling it — the ambiguity persists
- Treating the definition as a one-time decision — definitions need a quarterly review cadence
What changes when the decision is made
The visible signal of success is meeting tone. Leadership meetings start with the decision rather than with the reconciliation. The argument shifts from 'is this number right?' to 'what should we do about it?' That shift is worth more than any dashboard improvement we have ever shipped.
“Three answers to the same question doesn't mean the data is broken. It means the company has not yet decided what the question is.”
The one-line takeaway
Multiple versions of the truth is a decision-rights problem, not a data engineering problem. Name the owner, lock the definition, label everything else as a non-canonical view, and the cost of the disagreement evaporates — usually faster than the technology work would have shipped.
Published October 8, 2025 · 10 min read
