KPI dashboard: fewer numbers, checked more often

A KPI dashboard is a single view of the small number of measures that tell you whether the business is going where you intended. The word doing the work is key: a key performance indicator is one you would act on. If a number moving would not change anything you do, it is a statistic, and statistics belong somewhere other than the dashboard.

The failure mode is uniform and easy to recognise. Someone builds a dashboard with twenty-six tiles because twenty-six things can be measured, everyone admires it for a fortnight, and then nobody opens it — because reading it takes ten minutes and produces no decision.

What belongs on a KPI dashboard

• Between five and nine measures. Beyond that, attention divides and nothing gets acted on.
• A mix of lagging and leading indicators: revenue tells you what happened, pipeline and enquiry volume tell you what is coming.
• A comparison for every number — last period, same period last year, or target. A figure alone carries almost no information.
• The direction of travel, ideally as a small trend rather than a single arrow.
• An owner per measure: the person who explains it when it moves.
• The date the data was refreshed, so nobody argues from a stale figure.
• A defined threshold for what counts as a problem, decided before the number goes red.

The definition of each measure matters more than its presentation. Active customer, qualified lead and completed order all sound unambiguous and rarely are — two people produce different figures because they count differently. Write the definition next to the number, in one sentence, and most of the arguments in the meeting disappear.

Leading indicators are the ones worth arguing about

Lagging measures — revenue, profit, churn last month — are easy to agree on and impossible to act on directly, because the period has closed. Leading measures are the ones you can still influence: enquiries this week, proposals out, average response time, capacity booked for next month. A dashboard made only of lagging numbers is a report card. One that pairs each with a leading measure is a control panel.

Choosing leading indicators is genuinely difficult, because you are asserting a causal link — that more of this now produces more of that later. Getting it wrong is common. Reviewing whether the link held is the part that almost nobody does, and it is where the value is.

Building one that survives

1. Write down the three questions the leadership actually asks each month.
2. Choose the fewest measures that answer them, and reject everything else however easy it is to collect.
3. Define each measure in one sentence, including what is excluded.
4. Name an owner and set a target or threshold for each.
5. Pick a refresh rhythm that matches the decision: weekly for operational measures, monthly for financial ones.
6. Automate the data collection you can, and accept manual entry for the rest rather than dropping the measure.
7. Put the dashboard in the meeting where decisions are made, and start the meeting with it.
8. Review the measure set every quarter — remove one before adding one.

Vanity metrics and the rest

Numbers that only go up — cumulative registrations, total customers ever, page views — feel encouraging and inform nothing, because they cannot get worse. If a measure has no realistic path to going down, it is decoration. The same applies to averages hiding a distribution: average response time looks fine while a tenth of customers wait three days, and the median or a percentile usually tells the truer story.

Where to build it

Ettex Sheets is the practical answer for a small company: the numbers in a spreadsheet with the definitions beside them, charts on a summary tab, and several people able to open the same file rather than mailing versions around. That is unglamorous and it covers the case properly at this size — the constraint is rarely the visualisation, it is that nobody agreed what a qualified lead was.

It is not a business intelligence tool. There are no live connectors pulling from your other systems, no scheduled refresh, no data warehouse, no drill-down from a chart into the underlying rows across sources. Data arrives by import or by hand. Past the point where that hurts — many sources, daily refresh, several teams — a dedicated BI product is the right purchase, and worth making deliberately rather than by building an unmaintainable spreadsheet.

Five to nine measures, each with a definition, an owner, a comparison and a threshold — and one removed before any new one is added.