There's a specific failure mode in KPI work that looks like rigour and isn't. It goes like this. A working group spends two weeks producing a "comprehensive KPI framework." Fifty-three metrics. Colour-coded by department. Reviewed monthly in a deck that nobody quite finishes reading. Six months later, ask any executive which five metrics the CEO actually watches and you'll get a shrug. The fifty-three are still being reported. They're just not being used to decide anything.
The problem isn't too few KPIs. The problem is that the KPIs aren't connected to a strategic decision. They're connected to a measurement habit.
The question that cuts through
For every KPI you're considering — keeping, adding, or pruning — ask exactly one question: if this number changes by 20% in either direction, what decision changes?
If the honest answer is "we'd investigate," the metric is diagnostic. It tells you something is happening; it doesn't tell you what to do about it. Diagnostic metrics are useful, but they don't belong on your strategic dashboard. They belong in an operational layer, available when something triggers concern.
If the answer is "we'd reallocate engineering capacity from initiative X to initiative Y," or "we'd accelerate this bet and pause that one," or "we'd reopen the pricing question" — that's a strategic metric. It drives a decision. That's the short list you're building.
If the answer is "nothing, really" — and you'd be surprised how many KPIs return that answer when somebody asks honestly — the metric isn't a metric. It's a reporting tradition. Retire it.
The test for strategic relevance
A KPI earns a place on the strategic dashboard if it passes three tests, all three.
It must be directly connected to a named strategic priority. Not adjacent to one. Not a plausible downstream indicator of one. Directly connected. "Net revenue retention" for the retention bet. "Time-to-first-value" for the product-led growth bet. If you can't name the bet the metric belongs to in one sentence, the metric doesn't belong on the strategic list — no matter how often it's been reported in the past.
It must have one named owner. Not a committee. Not "the revenue team." A single person whose job it is to move that number this quarter, and whose performance review will reflect whether it moved. Metrics without owners are thermometers without doctors — interesting, occasionally alarming, attached to no treatment.
It must have structured, traceable work currently moving it. If there is no active work item explicitly linked to the metric, the metric is aspirational. Aspirational metrics generate beautiful charts. They don't generate results, and over time they generate cynicism, because everyone on the team can see that the number is being watched but nothing is being done about it.
Why most organisations have too many KPIs
The standard KPI process quietly conflates measurement with accountability. Finance measures revenue. Marketing measures pipeline. Product measures activation. Customer success measures NPS. Each function has its own dashboard, its own definition, its own quarterly read-out. The executive review aggregates everything into a master deck.
Nobody asks the question that would clean all of this up in one meeting: of these forty metrics, which three are the ones we are actually betting the company's strategy on this year?
The proliferation isn't malicious. It's the natural result of a process that adds without ever subtracting. Every quarter, someone proposes a new metric. Nobody retires the ones that have quietly stopped being strategic. The dashboard grows. The signal degrades. Forty metrics don't give you forty times the visibility. They give you forty times the noise, and a leadership team that has stopped trusting the dashboard because it can't tell which numbers actually matter.
The three-layer model
A clean KPI architecture has three layers, and they shouldn't be confused with each other.
Layer one is the strategic KPIs: three to five at most. These are the metrics that directly measure whether the strategy is working. They go to the board. They have single named owners. There is active, traceable work behind each of them. They don't change mid-year unless the strategy itself changes.
Layer two is the leading indicators: five to ten. These are the metrics that predict where the strategic KPIs are going next. If the strategic KPI is net revenue retention, the leading indicators might be time-to-first-value, support ticket density per account, or feature adoption rate among expansion candidates. These are reviewed weekly by the teams responsible for them, not monthly by the exec team.
Layer three is everything else: diagnostic metrics. Available, instrumented, queryable, but not on the agenda unless a leading indicator triggers a concern that requires them. Not reported on cadence. Investigated on demand.
The retirement conversation nobody has
Every quarter, before adding a single new metric, hold a ten-minute retirement review. Which KPIs on the current list no longer pass the three tests? Which ones got promoted to strategic at some point and never demoted? Which ones have been amber for three quarters with no structural response — which means the org has implicitly decided not to act on them, and they should either be moved to diagnostic or owned and acted on properly?
Retire them explicitly. Name what they used to track and why they no longer warrant a place on the strategic list. The act of retiring a metric is strategy work, exactly as much as the act of adding one is.
The Vindaris view
KPIs don't get aligned to strategy by being visible on a dashboard. They get aligned by being structurally connected — to the strategic priority they belong to, to the named owner accountable for moving them, and to the capacity that's been committed to do so. A metric without traceable work beneath it is not a strategic KPI. It's a hope with a number painted on it.
The dashboard isn't the strategy. The dashboard is supposed to be the surface of the strategy. When those two things are the same in your system, the question of how many KPIs you have stops being interesting — because the system will only let you have the ones that are real.