Too many KPIs is the same as none

A KPI only works if somebody changes their behaviour when it moves, and human attention caps out at roughly a dozen numbers watched properly. Past that point, every added metric dilutes the rest, and a forty-KPI scorecard delivers the same steering input as no scorecard at all — zero — while costing considerably more to maintain. The fix is a deliberate cut: a short list of decision-linked numbers, a graveyard for the rest, and a rule for how metrics earn their way back.

Why does more measurement produce less management?

Because measurement and attention are different resources, and only one of them scales. Software will happily compute two hundred metrics; the MD still has one working memory and one Monday morning. When the scorecard outgrows attention, people cope predictably: they skim, they watch the metrics that flatter them, and they stop noticing movement in any single number because every week something somewhere is red. Alarm fatigue arrives, and the scorecard becomes wallpaper.

There is a second failure underneath: fear of choosing. A forty-KPI report is often a decision deferred — nobody was willing to say which five numbers actually describe the business, so everything got included. The result reports everything and emphasises nothing. The discipline of choosing is the core of The MD Dashboard Blueprint: a dozen or so numbers, each tied to a decision, each with an owner, compiled by the system rather than by a person.

What does a KPI have to earn its place?

Four tests, applied without sentiment:

1. A decision. Name the action taken when this number moves adversely. No nameable action, no place on the list. "Interesting" is not a decision.
2. An owner. One person answers for the number and can actually influence it. Metrics owned by everyone are owned by no one.
3. A threshold. The level at which the owner acts, agreed in advance. A number without a trigger line is scenery.
4. Automatic compilation. When a metric requires manual assembly each week, then its cost is recurring and its accuracy is negotiable; it must come from the system of record or be redesigned until it can.

Applied honestly, these tests cut most scorecards by two-thirds. What survives is a working instrument panel; a daily-glance version of the same idea is set out in six numbers to check daily.

How do you run the cull?

The mechanism is a one-afternoon exercise plus a standing rule:

1. List every number currently reported. Dashboards, weekly emails, meeting slides — all of it, in one column. The length of this list is usually the first shock.
2. Apply the four tests. When a metric has a decision, an owner, a threshold and automatic compilation, then it stays; when it fails any test, then it moves to the graveyard — visible, archived, not reported.
3. Cap the survivors. When the surviving list still exceeds roughly a dozen, then rank by the money at stake in each metric's decision and cut from the bottom. The cap is the point; a cap that flexes is not a cap.
4. Install the entry rule. When someone proposes a new KPI, then it must name the decision, the owner, the threshold — and, if the list is full, which existing metric it replaces. One in, one out.
5. Revisit quarterly. When the business changes shape, then the list changes with it. A metric that earned its place last year can retire honourably.

Which numbers usually survive?

For a 5–50-staff B2B service firm, the survivors cluster reliably: deals created, weighted pipeline against target, conversion at the proposal boundary, revenue against plan, and cost per client. Firms with retainer income add the recurring-revenue view — worth measuring properly even in a project-based firm, as set out in tracking recurring revenue in a project-based firm, because it is the number that describes how much of next quarter is already sold. And one slower number belongs on every list: client lifetime value, the figure that tells you what a client is actually worth and therefore what you can afford to spend winning one — the arithmetic behind lifetime value: the number that justifies the system.

The exact survivors vary by sub-vertical — a recruiter's pipeline arithmetic is not an MSP's, and the sector-by-sector differences are mapped in Growth Systems by Industry — but the count barely moves. Around a dozen, whatever the business.

Doesn't cutting KPIs mean flying blind?

The opposite, in practice. The graveyard metrics still exist in the CRM and can be pulled for diagnosis whenever a headline number misbehaves — nothing is deleted, it is simply demoted from "watched weekly" to "available on demand". What the cut removes is not information but noise, and noise was the thing causing blindness. A dozen numbers watched with real attention, each wired to a decision, will catch a problem months before a forty-metric scorecard that everyone has learned to scroll past.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks, including which numbers deserve your attention, and what to build first.