Skip to content

Technology | People | Progress A Brighter Tomorrow

Get in Touch
Abstract graphic of a chart panel in the Logic Networks colours, introducing the article on building a first dashboard

November 17, 2026

A first dashboard that a board will read

Most first dashboards fail the same way. They show everything the systems can produce, arranged attractively, and after the first meeting nobody opens them again.

The ones that survive are smaller, duller and answer a question somebody was already asking.

Start from the argument, not the data

Before opening any tool, write down the three questions the board actually argues about. Not the ones it should ask. The ones that come up every time.

They are usually simple. Are we selling enough. Are we collecting the money. Are we delivering on time. Are people leaving.

A dashboard that answers those gets read. One built from whatever the finance system happened to export does not, however good it looks.

Six numbers, on one screen

Six is enough for a first version and more than most organisations can maintain honestly. If it does not fit on one screen without scrolling, it is a report rather than a dashboard, and reports get filed.

Resist the request to add a seventh. Adding is easy and it is how you end up back at the version nobody reads. Keep a list of the things people asked for and revisit it after three months.

Four cards: Six numbers not sixty, A direction on each, One owner per number, A stated source

Every number needs a direction

A figure on its own means nothing. Sales of a given amount this month is not information until you know what it was last month and what it was this month last year.

Put both beside it. For anything seasonal, the same month last year matters more than last month, and putting the wrong comparison next to a number is worse than putting none.

Avoid targets on a first version unless the target is already agreed. Arguing about the target is a good conversation and it is not the one you want at the first showing.

Every number needs an owner

A named person who can explain a change without a week of preparation. If a number moves and nobody can say why within a day, it is not ready to be on the screen.

This is the check that removes most of the numbers people want to include. It is also the reason dashboards improve the underlying data: once a person’s name is attached, the quality gets attention.

Say where each number came from

The source system, the definition, and when it was last refreshed. A small line under each figure is enough.

The first serious challenge to any dashboard is somebody saying the number is wrong. Usually it is not wrong, it is defined differently from the way they define it in their head. Having the definition written down turns a long argument into a short one.

Show the refresh time as well. A number that is three days old is fine as long as everybody knows it is three days old.

Expect the data to be worse than you think

The first attempt at any dashboard finds duplicates, blank fields, twelve spellings of the same category and dates in three formats. That discovery is the most valuable part of the exercise, and it will take longer than building the screen.

Do not hide it. A number with a note saying a proportion of records are missing a category is honest and useful. A number that silently ignores them is neither.

Decide how often it refreshes, and say so

A dashboard that updates constantly invites people to watch it, which is rarely useful for the kind of numbers a board looks at. One that updates monthly invites the question of whether it is current.

Match the refresh to the decision. Numbers a board reviews monthly can refresh monthly. Numbers somebody acts on weekly should refresh weekly. Very little in a first dashboard needs to be live, and live data costs more to build and more to keep working.

Then put the refresh time on the screen. It stops the only question nobody can answer in a meeting, which is how old the figures are.

Build the first one quickly and expect to throw it away

A rough version in front of people in a fortnight beats a polished one in six months, because the feedback is where the real requirements come from.

Say at the outset that version one is a draft. It gives everybody permission to criticise it, which is what you want.

Then keep it for a quarter before adding anything. The questions people ask while using it are a better guide than the ones they asked before it existed.

We build these with the organisations we support, usually starting from data they already have. There is more on our data and AI page, and the education version is on the education page. If you have a board pack that takes two days to assemble each month, tell us what is in it and we will say what we would automate first.