Choosing what belongs on a dashboard

A dashboard built to answer every question answers none of them. The eye lands nowhere, and after a fortnight nobody opens it.

The useful version holds few numbers, chosen because somebody would act differently depending on what they say.

The test is whether it changes a decision

A number nobody would act on is a number that belongs in a report somebody runs occasionally, not on a screen everybody passes.

Total revenue this year is interesting. Jobs finished but not invoiced is actionable, because somebody can do something about it this afternoon.

Name the reader before the metric

An owner, a site supervisor and a bookkeeper need different screens. One dashboard serving all three serves none of them well.

Deciding who reads a board settles most arguments about what goes on it, and it usually shortens the list considerably.

Leading numbers beat trailing ones

Revenue is a result of decisions made weeks ago. Quotes sent, quotes accepted and jobs booked predict it while there is still time to change the outcome.

A board of trailing numbers reports history accurately and helps nobody plan.

Every number needs a comparison

A figure on its own carries no meaning. Forty thousand is good or bad only against last month, last year or a target.

The comparison is what makes a glance sufficient. Without it, the reader has to remember what normal looks like, and they will not.

Say what the reader should do

A red number with no obvious next step produces anxiety rather than action. The board should point somewhere.

Overdue invoices should open the list of them, and unassigned jobs should open the schedule. A number nobody can click becomes a number nobody acts on it.

Five is usually the right count

A screen with five numbers gets read properly. One with fifteen gets scanned, and the important one hides among the rest.

Anything that fails the decision test moves to a report. Our guide to a home screen covers building the short version well.

Retire what nobody reads

Dashboards accumulate. Somebody asks for a metric during a bad quarter and it stays on the board for three years.

Asking once a year which numbers actually changed a decision removes more than it adds, which is the point.

A wrong number is worse than no number

A figure people have learned to distrust poisons the whole screen, because nobody knows which of the others to believe.

Checking each metric against the underlying records once, when it is built, is what buys the trust that makes a dashboard useful.

Define the number once, in writing

Two people reading jobs completed can mean different things, and both will defend their figure.

A one line definition beside each metric settles that permanently, and it is the first thing anybody asks when a number looks wrong.

Place the urgent ones where the eye lands

The eye starts at the top left, so the most consequential number belongs there rather than wherever somebody added it.

Grouping the rest by who acts on them keeps a shared board usable without splitting it into three.

A board is not a report

Month end analysis, tax summaries and anything needing a date range belong in a report somebody runs deliberately.

Putting them on a dashboard makes the screen slower and the important numbers harder to find.

A dashboard reading live data also has a cost in speed. A screen that takes ten seconds to load is a screen people stop opening, whatever is on it.

What to change first

• Every number passes the test of changing a decision
• The reader gets named before the metric gets chosen
• Leading numbers sit above trailing ones
• Each figure carries a comparison beside it
• Every number opens the list behind it
• One yearly review retires whatever nobody read