Your KPI Is Green. Here's the Question Your Board Isn't Asking.
For CEOs, CFOs and CTOs
Metric Audit · Business AnalyticsRevenue is on track. Uptime is on track. Customer satisfaction is on track.
And yet the business doesn't feel like it's improving as fast as the dashboard says. The gap is rarely bad data. It's that "on track" is a comparison, and almost nobody in the room asks what it's being compared against.
Every executive team has seen a slide full of green that didn't match what the people closest to the work were saying. This piece is about that gap: how a KPI can improve without the business improving, and three questions that expose it before your next board meeting.
1. The Meeting Where Two Numbers Disagreed
Quarterly review. Sales reports revenue growth of 12%. Finance reports 9%. Both numbers came from the same CRM, exported the same morning.
Nobody made an error. Sales counts a deal as closed when the customer signs. Finance counts it when the order is booked and can be invoiced. The three points between the two figures are deals signed in the quarter but not yet booked — real revenue on a real timeline, which one team's definition counts and the other's doesn't.
The next half hour goes to reconciling the two figures instead of deciding what to do about either of them.
The dashboard didn't fail. The word underneath it did.
This isn't a rare failure. It's what happens by default in any organization where a metric's name is shared and its definition isn't.
2. Four Ways a KPI Goes Green Without the Business Improving
The two-numbers meeting is the visible version of the problem. The quieter version is a single number that everyone agrees on — and that has stopped meaning what they think it means.
| The target moved | The goal was lowered or re-baselined to reflect current reality, so the same performance now reads as "on track." |
| The definition drifted | What counts as "resolved," "active," or "qualified" was loosened, or is applied differently by different teams. |
| The population changed | Exceptions, exclusions, or "out of scope" categories grew, so the average improves while the underlying work doesn't. |
| The metric can't fail | There is no stated threshold at which the number would trigger action, so it can only ever report reassurance. |
Each of these is invisible on the dashboard itself. The number is green, the chart trends the right way, and the only place the problem shows up is in the conversation between the people who run the work and the people who read the report.
The same pattern appears well beyond finance. In IT operations, a resolution-time metric can improve for quarters while the same incidents keep coming back. In knowledge management, article counts and view counts look healthy while ticket volume doesn't move. In each case the number is accurate. It's the meaning that has drifted.
3. Metric Drift, and the Watermelon KPI
The gradual change in what a KPI actually measures — through a redefined term, a re-baselined target, or a shifting population — without any visible change to the metric's name or the dashboard that reports it.
Management writing has a name for the extreme version: the watermelon KPI, green on the outside and red on the inside. It isn't usually the result of anyone hiding anything. It's what accumulates when reporting is optimized for looking on track and nobody is accountable for whether the measure still reflects reality.
It isn't a new observation, either. Goodhart's law states it plainly: once a measure becomes a target, it stops being a reliable measure. The moment a number is tied to a goal, a bonus, or a board expectation, every decision around it — how to define it, where to draw its boundaries, when to reset it — becomes a decision about the result as well as the method.
4. Why Your Current Reporting Process Doesn't Catch This
Most reporting governance is built to answer questions like these: Is the data complete and accurate? Did the dashboard refresh on schedule? Does the calculation match the specification? These are necessary controls, but they test whether the number was calculated correctly, not whether it still means what the board believes it means.
A metric can pass every one of those checks and still be measuring the wrong thing. The questions that catch drift are different — and they're business questions, not technical ones.
- When was this metric — or its target — last redefined, and by whom?
- Who owns the definition — not just the number?
- What result would this metric need to show before someone acted?
Run these against your top three board-level metrics. If you can't answer all three quickly for each of them, that isn't a data problem. It's a governance gap — and it's cheaper to find it in a review meeting than in a restatement.
5. A 30-Minute Exercise: Trace the Split
You don't need a tooling project to start. Take the most recent number that two teams disagreed about, and do this:
- Write both calculations in plain language, side by side. Not the formulas — the sentences. "Revenue is deals signed in the period." "Revenue is orders booked in the period."
- Find the first line where they differ. It's almost always a single word: active, closed, qualified, resolved, churned.
- Ask who decided what that word means in each team, and when. If nobody can say, you've found an orphaned definition.
- Write one definition, name one owner, and record the date. The aim isn't to force every team onto one number for every purpose. It's that anyone reading a number knows which definition they're looking at.
That last step matters more than it looks. Finance and Sales may legitimately need different views. What they can't afford is for both to be called "revenue" on the same slide.
6. What Governed Metrics Look Like
One definition, one named owner
Every board-level metric has a written definition and a person accountable for the definition itself, not just for producing the number. "The BI team" is not an owner.
Targets carry a change log
When a target or definition changes, the date, reason, and approver are recorded, and the board can see that it changed.
Every metric has a failure condition
A stated threshold triggers a specific action. If a number has never prompted a hard conversation, ask whether it can.
Speed is paired with quality
Resolution time sits next to recurrence. Volume sits next to outcome. A faster number that hides a worse one gets caught.
None of this slows reporting down. It changes what the report is allowed to claim.
7. The 5-Question Self-Assessment
Run these against your current board pack. Three or more unclear answers means your reporting is scaling faster than your definitions.
| # | Question | Yes / No / Unsure |
|---|---|---|
| 1 | Can you name the person accountable for the definition of your most-quoted board metric? | |
| 2 | Do you know the date each headline target was last changed, and why? | |
| 3 | If two teams reported different numbers for the same metric this quarter, could you show where the calculations diverge within a day? | |
| 4 | Does every headline metric have a stated threshold that would trigger action? | |
| 5 | Is each speed or volume metric paired with a quality or outcome metric on the same page? |
8. Three Questions We Hear Most Often
We already have a data governance program. Doesn't that cover this?
Data governance usually covers quality, lineage, and access — whether the data is right and who can see it. A metric definition is a business decision layered on top of that. A perfectly accurate calculation can still measure the wrong thing, so most organizations don't need to replace their data governance. They need metric-level ownership on top of it.
Won't locking down definitions make reporting rigid?
Definitions should change when the business changes. The goal isn't to freeze them; it's to make every change visible. A change log means the board can see that a target moved and why, which is what turns a quiet adjustment into a deliberate decision.
How does this matter for AI and automated reporting?
Any AI assistant or automated report that answers questions from your dashboards inherits your definitions. If two teams define "active customer" differently, it will confidently give either answer. It's the same question we raised in who owns the risk when AI acts: everyone can explain their part, and nobody owns the outcome.
The Question for Your Next Board Meeting
Before the next board pack goes out, pick the metric everyone trusts most and ask one thing: what would have to be true for this number to be green while the business got worse? If the answer takes more than a minute to give, the number is telling you less than you think.
Facing similar ServiceNow challenges?
Get a free ServiceNow assessment and understand where your platform may be losing value.
No spam. No sales pressure. Only actionable insights.