KPIs do not create
performance

Almost every organisation we meet has metrics. Many have far too many. Very few can say, metric by metric, which decision is triggered when one turns red. That is exactly where performance is lost — not in the measuring, but in what comes after it.

The measurement paradox

Organisations have never had so many metrics available. Performance has not followed. According to McKinsey's The State of Organizations 2026, fewer than 25% of organisations achieve sustained performance improvement — and that is despite measurement systems that are more complete and more automated than ever.

The pressure, meanwhile, is not easing. In the same study, 43% of executives rank productivity as their top priority, and 61% report strong pressure to deliver gains. In other words: we have never measured so much, never expected so much in return — and the rate of genuine transformation remains low.

The chain stops before the decision

A metric follows almost exactly the same path in every organisation: it is measured, it is reported, it is discussed in a meeting — and there, very often, the chain stops. The fourth step, deciding, has no mechanism behind it. Nobody defined in advance what should happen if the number moved outside its acceptable range.

The result is an organisation that looks like it is steering: polished dashboards, regular reviews, numbers discussed seriously. But between the discussion and the action there is no mechanism — only good intentions. Harvard Business Review puts the test as bluntly as it can be put: a metric that cannot change a decision is not a steering tool, it is one more line of reporting.

This follows directly from our article on meetings and steering: a well-run ritual guarantees nothing about what comes out of it. And from the one on executive committee visibility: having a reliable number is worthless if it triggers no trade-off.

Four conditions for a metric that actually steers

1. A named owner

A specific person, not a department. "Finance tracks that number" is not an owner — it is a diffuse zone of responsibility. If nobody is named, nobody genuinely watches, and the metric only becomes visible once it is already too late.

2. A fixed review date

A slot already in the calendar, never "when we get the time". The frequency matters less than the fact that it is actually settled: a metric reviewed monthly on a fixed date steers better than one theoretically monitored continuously and looked at twice a year.

3. A threshold set in advance

The level that triggers the alert is fixed beforehand, not negotiated once red is reached. This is the most frequently skipped condition — and the most protective: when a threshold is debated at the moment it is crossed, it is almost always revised upwards rather than acted upon.

4. A pre-committed decision

You already know what you will do if the metric turns red: reallocate, stop, escalate, reinforce. Not necessarily in detail — but the nature of the response is settled in advance. Without that, the metric is decorative, however carefully it is produced.

The three-question test

This is the highest-return exercise an executive committee can run, and it takes no more than an hour. Take your current dashboard, and put every metric through the following test:

1. Who? Which person, by name, looks at this metric — not which department?
2. When? On what precise date is it reviewed, and is that slot already in the calendar?
3. So what? Which decision is triggered if it turns red? If the answer is "we will discuss it", there is no decision.

A metric that fails even one of these three questions steers nothing. It has two possible outcomes: delete it, or repair it. In our experience, most executive dashboards come out of this exercise considerably lighter — and considerably more useful.

The gain is not cosmetic. A dashboard reduced to the metrics that genuinely trigger action makes reviews shorter, trade-offs faster, and above all makes visible what volume had been hiding: the few signals that actually matter.