How to design a dashboard people actually use

The screen nobody opens

There is a kind of project that ends badly despite going well: the data loads by itself, the figures reconcile with the source system, the screen holds 34 charts and plenty of color. Three months later the usage stats say it has been opened four times.

The reason is almost always the same: it was designed from the data that happened to be available rather than from the decisions someone has to make. A dashboard is not a display cabinet for information; it is a working tool that has to make somebody do something differently today.

Start from the decision

Before drawing anything, fill in this brief together with the people who will use the screen. For every decision you need to know who makes it, how often, what triggers it and what action follows. If any of that cannot be completed, the metric probably does not belong there.

Restock an item — made by the store manager, daily, when cover drops below 5 days. Action: raise a warehouse order.
Stop a line — made by the shift supervisor, every shift, when the scrap rate goes above 3%. Action: check tooling.
Review pricing for a family — made by the sales director, monthly, when margin sits two points below target. Action: new price proposal.

That brief is the design. Everything else is execution. When teams run this exercise, the initial list of 25 "essential" metrics usually shrinks to 8, and two or three nobody had asked for turn out to be the ones that actually trigger decisions.

Three levels, not one

A useful screen works at three different speeds.

Five seconds. Is it going well or not? Three or four large figures at the top, each with its comparison: against target, against the same period last year, or against last week. A number without a reference says nothing: €412,000 can be great news or a disaster.

Two minutes. Why? This is where trends over time and comparisons by category, store, line or customer belong. It is the area where users look for the cause of a deviation.

The detail. When needed, you drill into the fine table, filter and export. This level should take up no room on the main screen: it needs to be one click away.

Visual hierarchy

The eye reads a screen the way it reads a page: top left first, then downwards. Use that.

The most important thing goes highest and largest. Size should track relevance, not the amount of data behind it.
One accent color for emphasis. If everything stands out, nothing does.
Meaningful color always means the same thing: if red is "off target" in one chart, it cannot be "northern stores" in the next.
Regular alignment and spacing. A clean grid scans far faster than a mosaic of misaligned cards.
Short labels in business language. "Gross margin", not "mrg_gr_calc".

Which chart for which question

Choosing a chart is not an aesthetic decision; it follows from the question.

How does this change over time? A line, with enough points for the trend to be visible.
Which categories stand out? Horizontal bars sorted from largest to smallest. Sorting does half the work.
How much does each part contribute? Stacked bars or a single segmented bar. Pie charts only work with two or three slices; with eight they are unreadable.
Where do cases concentrate? A histogram, or a heatmap by hour and weekday.
A single figure against a target? One large number with the variance beside it and, if there is room, a small line showing the last twelve periods.
Is there a relationship between two quantities? A scatter plot.

Three things to avoid without exception: 3D effects, dual axes with different scales, and bar charts whose vertical axis does not start at zero. All three exaggerate differences that are not there.

What to leave out

Metrics nobody looks at but that "look good".
All the detail "in case someone needs it". If they need it, let them drill into it.
Decorative headers, oversized logos and welcome messages that take up half the screen.
More than eight or nine visual objects per screen. Beyond that, people stop reading and start hunting.
Filters left wide open by default. Let the screen start on the most useful view already applied.

Details that decide adoption

Four dull things explain much of the difference between a screen that gets used and one that does not:

Speed. If it takes more than three seconds to load, people stop checking it on impulse.
Mobile. A store manager or a shift supervisor will never open a laptop to check one number. They need a vertical version with the three key metrics.
A visible refresh timestamp. One small line saying when the data was last loaded removes half the questions and all the distrust.
Accessibility. Sufficient contrast, readable text sizes and no information conveyed by color alone.

Test it with five people

Before publishing, sit with five real users, one at a time. Give them a specific question ("which store dropped the most this week?") and watch without saying anything. Time them.

If it takes more than thirty seconds to find, the structure is not clear.
If they ask what a label means, change the label.
If they go straight to the filter and change it, the default view is wrong.
If they drop into the detail table immediately, the summary probably is not answering the question.

Five fifteen-minute sessions catch nearly every design problem. It is the highest-return investment in the whole project.

The first two weeks

Adoption is won right after publication:

Set a retirement date for the old report and announce it. While both coexist, the old one wins.
Run the first weekly meeting with the screen on the projector, so everyone sees it is the official source.
Thirty minutes of training, no more. If it needs two hours, the design is the problem.
After a fortnight, collect three concrete improvements and ship them quickly. That gesture does more for adoption than any new feature.

Checklist before publishing

Every element answers a question somebody genuinely asks.
Every figure carries a comparison.
The data reconciles with the source system for a closed period.
The screen reads well in light and dark mode, and in portrait.
The last refresh date is visible.
There is a path from summary to detail without leaving the screen.
Permissions are configured by role.
Someone outside the team understood it with no explanation.

If your current screen does not pass this checklist and you want a design review, write to me. I run a free diagnostic and tell you what I would change and why.