What belongs on the screen.

Written by the Incognito Labs engineering team 16 September 2026 8 min read

Nearly every business that commissions a dashboard is enthusiastic about it for about three weeks. Then it quietly stops being opened, and a year later somebody asks whether anyone still uses that thing we paid for.

It is almost never because the software was bad. It is one of three things, and all three are decisions made before any software was chosen:

• It showed what was easy to measure rather than what anyone acts on. Whatever the tools exported became the dashboard, because that was the path of least resistance.
• Nobody could do anything differently as a result of it. Interesting is not the same as useful. A number that changes no decision is decoration.
• It went stale. One figure was wrong once, nobody could explain why, and from that day the whole screen was treated as approximate.

That third one is fatal and it is worth being blunt about: a dashboard has exactly one job, which is to be trusted. It does not recover from being wrong in public.

Before a figure earns space on the screen, it should survive one question:

This one test removes most of what people initially ask for. It is also unpopular, because the figures it removes tend to be the flattering ones. A dashboard that passes it is smaller, plainer and considerably more useful than the one most people imagine at the start.

A working rule of thumb: if you cannot take in the whole screen in about ten seconds, it is a report, not a dashboard. Reports are fine. They are just a different thing, read at a different cadence, and they should not be competing for the same space.

Across every business we have built one for, the useful content falls into four groups. Most screens need all four and nothing else.

1. Position — where we stand right now. Revenue this month, jobs open, cash in. The handful of figures you would quote if someone rang and asked how business is. These are the ones people expect, and on their own they are the least useful of the four.

2. Direction — which way it is moving. A number with no comparison is barely information. £70,000 means nothing; £70,000 against £64,000 last month means something, and against £81,000 means something else entirely. Always show the comparison next to the figure, never on a separate tab.

3. Exceptions — what needs a person today. This is the part most dashboards miss and the part that earns the daily visit. Overdue jobs, unpaid invoices past terms, enquiries not yet answered. Not a count — the actual list, clickable, so noticing and acting are the same gesture.

4. Commitments — what we have promised. Work due this week, deliveries expected, appointments booked. Position tells you where you are; commitments tell you what is already spent. A business can look healthy on position and be in trouble on commitments.

Image placeholder Suggested: an annotated dashboard screen with the four zones labelled — position, direction, exceptions, commitments — in the lime accent.

Some figures are not merely useless on a dashboard, they are actively harmful, because they are reassuring and wrong at the same time.

• Totals since the beginning of time. Cumulative figures only ever go up. A chart that cannot fall is a chart that cannot warn you.
• Averages without a spread. An average job value of £400 looks entirely different if it is forty jobs at £400 or two at £7,000 and thirty-eight at £50. Show the distribution, or at least the median alongside.
• Percentages on small numbers. “Conversion up 50%” is two jobs instead of one. Always show the underlying count next to the percentage.
• Activity mistaken for progress. Calls made, emails sent, tasks ticked. These measure effort, and effort is an input. Put them on a team screen if useful, never on the screen that tells you how the business is doing.

The instinct is to build one dashboard everyone shares. It produces a screen that serves nobody: too much detail for the owner, too much money for the employee, the wrong things entirely for an outside partner.

The same underlying data should be shown four different ways. An owner needs margin, direction and exceptions. An operations lead needs today’s commitments and what is at risk. An employee needs their own work and nothing about anyone’s pay. An outside partner or client needs their jobs only — no margin, no other customers.

This is not only a presentation question, it is a governance one. “Who can see the payroll figure” should be answered by the system, not by remembering not to share a tab. If your current answer involves keeping a separate copy for the subcontractor, that is the spreadsheet problem described in seven signs your business has outgrown its spreadsheets, and no amount of charting will fix it.

The dashboard demo in our services section is built around exactly this — the same business, four seats, each showing what that seat is entitled to.

Given a choice between a beautiful dashboard updated on Mondays and a plain one that is right now, take the plain one every time. The moment a screen needs someone to refresh it, two things happen: it is out of date more often than not, and nobody is sure how out of date, which is worse.

Two practical rules we apply to every build:

• Put the time on the screen. “As at 09:14” in the corner. Cheap to add, and it converts “is this current?” from a doubt into a fact.
• Show a gap as a gap. If a source has not reported, say so rather than drawing a zero. A zero is a claim, and an incorrect one costs you the whole screen’s credibility.

Getting to live figures usually means the underlying data has to arrive without anyone carrying it — which is a plumbing problem before it is a charting one, and the reason reporting comes third in the order we set out in what to automate first. If you are weighing whether to build this or configure an off-the-shelf tool, the arithmetic is in custom software or off-the-shelf.

Write down every question you currently ask someone else to answer about the business. Next to each, write what you would do differently if the answer were bad. Cross out every line where that second column is empty.

What remains is your dashboard, and it will be shorter than you expected. Sort it into the four groups, decide who is allowed to see each line, and you have a specification — without having chosen a single piece of software.