Indicators that actually get used: designing to decide

An indicator nobody uses is not a visualisation problem. It is an indicator tied to no decision. Before designing the dashboard it is worth answering what will be done differently depending on what it shows.

What follows: why dashboards get abandoned, what makes an indicator actionable, what requires judgement, and how to sequence the exercise.

Almost every organisation that has invested in analytics has dashboards nobody opens any more. They were built with effort, presented with enthusiasm, and went quiet without anyone noticing.

The usual explanation is that adoption or training was lacking. The real explanation is usually simpler: the dashboard showed correct information that changed no decision.

Why dashboards get abandoned

The first reason is the absence of a recipient. A dashboard built “for management” has no owner; one built for whoever decides stock replenishment does.

The second is the lack of a threshold. A number without a reference does not say whether things are good or bad, so whoever looks at it has to supply the judgement every time, and that wears thin.

The third is overload. Twenty indicators on one screen communicate that none is a priority, and the reader ends up looking at the one they already knew.

What makes an indicator actionable

It is tied to a concrete decision. Somebody exists who, depending on its value, does something different. If nobody changes behaviour on the number, it is context, not a management indicator.

It has a threshold and an owner. It is known from which value action is required and who acts. Without those two things the indicator informs but does not direct.

It can be influenced. Measuring something the recipient has no lever over produces frustration, not management. The exchange rate affects the result and is not a management indicator for the sales team.

It has a single, published definition. If two areas calculate it differently, the committee’s time goes on reconciling figures instead of deciding.

What requires judgement

Choosing few. The discipline of limiting the dashboard to the indicators that direct the operation is uncomfortable, because it means leaving out interesting things. It is also what makes it get used.

Telling an outcome indicator from a leading one. The outcome arrives too late to correct; the leading one allows action while there is still time. A useful dashboard combines both and says which is which.

And accepting that some indicators must be retired. A dashboard that only grows ends in the overload that switches it off; reviewing it periodically and removing what no longer directs anything is part of the maintenance.

How to sequence the exercise

Start from the decision, not from the available data. Sit with the recipient and ask which decisions they take in their week and what information they lack to take them with more confidence.

Translate each decision into an indicator with a threshold, an owner and an expected action. Write it in one line: when this value crosses this limit, this person does this.

Only then check whether the data exists. The reverse order — looking at what data is available and building indicators from it — is what produces dashboards that are correct and sterile.

What changes by level

Operations need frequency and detail: few indicators, refreshed often, with the ability to drill to the individual case in order to act.

The board needs the opposite: aggregation, trend and comparison against target, with the option to ask for the detail when something stands out.

Serving both with the same dashboard usually ends up serving both badly. They are worth designing separately even when they share the same underlying definition.

What has to exist first

A recipient identified by name and an agreement on the definition of each indicator. Without the first there is nobody to ask whether it helps; without the second, every review starts by arguing about the figure.

It is also worth fixing at the outset how it will be known whether the dashboard works. The most honest measure is simple: how often it is opened, and how many decisions are attributed to it.

A format that works

The simplest way to confirm an indicator is well framed is to force it into a single sentence with five elements: what is measured, for whom, at what threshold, who acts and what they do.

For example: the percentage of requests served outside the committed deadline, for the operations owner, with an alert above the agreed limit, who reviews the team’s workload allocation that same week.

When the sentence cannot be completed, the same thing is usually missing: nobody holds the lever to act, or there is no agreement on the threshold. Both are useful findings and neither is solved by building the dashboard.

The format has a practical advantage too: it becomes the indicator’s published definition directly, so whoever consults it months later finds the criterion next to the number.

How an indicator is retired

Retiring matters as much as adding, and it costs more because nobody wants to be the one who switches something off. It is worth treating as a periodic review rather than an individual decision.

The criterion can be explicit: an indicator not consulted within an agreed period, or whose owner cannot recall taking any decision with it, leaves the main dashboard and moves to a secondary view before being removed.

How many indicators should a dashboard have?

The ones that direct its recipient’s decisions, and they are usually few. When the screen holds twenty, the reader concludes none is a priority.

How do you know whether an indicator is useful?

By asking what is done differently depending on its value. If the answer is nothing, it is context information and worth treating as such.

Is one dashboard for operations and the board a good idea?

Usually not. Operations need detail and frequency; the board needs aggregation and trend. One dashboard for both tends to serve both badly.

What should be done with dashboards nobody opens?

Retire them, after confirming with the recipient that they direct no decision. Keeping dead dashboards costs credibility and support effort.