Decision First
What this screen is for, before which charts exist.
Dashboard UX Design
Dashboard UX Design That Shows What Needs Attention
Most dashboards are built from what the data warehouse can produce rather than from what anyone needs to decide, so they become inventories of available metrics. Dashboard UX design starts from the opposite end: what decision is this screen for, what would change it, and what can therefore be left off.
We Start From the Decision
A dashboard nobody acts on is a report with a refresh button.
Attention Directed
What is abnormal, surfaced — not twelve equal tiles.
Comparison Built In
A number without a baseline cannot be interpreted.
Drill-Down Designed
From "something is wrong" to "here is what".
Fewer Metrics
Removing charts is usually what makes a dashboard useful.
What a dashboard is judged on
A dashboard exists to support a decision. If someone looks at it and does not know what to do differently, it has failed regardless of how much data it displays. These checks test that rather than visual sophistication.
Decision mappingWhich specific decision each view exists to support
Time to answerHow long it takes to answer the question the dashboard was built for
Default view relevanceWhether the first screen shows what most users need most often
Metric definitionsWhether every figure has a stated definition users can check
Comparison contextWhether numbers appear alongside a baseline that makes them meaningful
Chart type fitWhether each visualisation suits the data rather than the layout
Exception visibilityWhether things needing attention are distinguishable at a glance
Colour independenceThat status is not conveyed by colour alone
Data freshnessWhether users can tell how current the figures are
Empty and partial statesWhat the view shows before data exists or when a source fails
Drill-down pathWhether a user can move from a number to the records behind it
Export and sharingWhether findings can leave the dashboard in a usable form
Metric definitions cause more disputes than any design decision. When two teams read the same figure differently, the dashboard stops being a shared reference and becomes something each department argues with. Definitions visible at the point of use resolve this permanently.
Dashboard UX, Explained
Why Do Dashboards Go Unused?
Four reasons, and all four are design decisions.
Discuss Your Dashboard →- 1
Everything Looks Equal
Twelve tiles of the same size, so nothing stands out.
- 2
No Baseline
A number with nothing to compare it to means nothing.
- 3
No Next Step
Something is wrong and the screen offers no way to investigate.
- 4
Built From Data
Assembled from what exists rather than what is needed.
What Dashboard UX Work Covers
The decision, the hierarchy, and the route from signal to cause.
For internal tools generally, see B2B UX design. For internal tools generally this sits alongside B2B UX design; query performance behind slow dashboards is database development.
How We Design Dashboards
Ask what would change as a result of looking. If nothing, the chart does not belong.
Define the Decision
What action this screen supports.
Select Metrics
Only those that would change the decision.
Build the Hierarchy
Prominence by consequence, not by tile grid.
Design Drill-Down
From anomaly to cause.
Remove
The charts nobody uses, which is usually most of them.
What the dashboard is actually for
Dashboards serve very different purposes and the purpose determines almost everything about the design. Building one view intended to serve all of them is the most common reason dashboards go unused.
Monitoring dashboards
Checked frequently and briefly to confirm nothing is wrong. Success is that a normal state can be verified in seconds and an abnormal one is impossible to miss. Density and stability matter; anything decorative competes with the signal.
Analytical dashboards
Used to investigate a question, which means the user needs to slice, filter and drill rather than read. Flexibility and drill-down matter more than immediate legibility, and the design should assume exploration rather than a single glance.
Reporting dashboards
Produced for someone who was not involved in the work and needs the conclusion. These need narrative framing, comparison against a baseline, and enough definition that a reader unfamiliar with the metrics does not misinterpret them.
Operational dashboards
Watched during a shift and used to decide what to work on next. Real-time behaviour, clear queueing and unmissable exception handling matter, and the design has to survive being viewed continuously without fatiguing the person watching it.
Customer-facing dashboards
Where the audience did not build the metrics and cannot ask what they mean. Definitions, honest handling of gaps, and restraint about what is shown all matter more here, since a confusing figure becomes a support ticket rather than an internal question.
Selection
Why More Metrics Make a Dashboard Worse
Every additional chart reduces the attention available for the ones that matter.
How do dashboards accumulate?
Someone asks for a metric, it is added, and it is never removed. Requests are cheap to grant and there is no cost visible at the moment of adding.
The cost is real and diffuse: each new tile takes a share of the screen and a share of the reader's attention, and the important numbers get harder to find.
A useful discipline is that adding a chart requires nominating one to remove. It forces the comparison that additive requests never do.
Why is a number without context useless?
Because nobody knows whether 847 is good. Interpretation requires a comparison — last week, last year, the target, the usual range.
A dashboard of bare figures makes every reader do that work from memory, which means most of them do it badly or not at all.
Showing the comparison alongside the number, and marking what is outside the normal range, is what turns a display of data into something someone can act on.
Most dashboards answer questions nobody asked
Dashboards are commonly built from what the data warehouse can produce rather than from what someone needs to decide. The result is a dense grid of accurate, well-rendered charts that nobody consults, because none of them corresponds to a decision anyone actually makes.
The corrective is to start from the decision. What does this person do differently depending on what they see, and how often do they make that choice? A metric that changes no behaviour, however interesting, belongs in an exploratory tool rather than on a dashboard someone is expected to check.
This usually reduces the number of visible metrics substantially, which is the point. A dashboard showing six things that drive decisions is used; one showing thirty is scanned once and abandoned, because the effort of locating the relevant number exceeds the value of knowing it.
A number without context is not information
A figure presented alone cannot be acted on, because the viewer has no way to know whether it is good. Four hundred of something is meaningless without knowing whether last month was three hundred or five hundred, and whether the target was either.
The context that makes a number actionable is usually one of three things: comparison to a previous period, comparison to a target, or comparison to a related segment. Adding one of these turns a display into a signal, and it costs almost no additional space — a small delta or a sparkline alongside the figure is generally enough.
Trend matters more than the current value in most operational contexts. A metric that is below target but improving steadily calls for a different response than one that is above target and falling, and a dashboard showing only the current figure cannot distinguish them.
Encoding status so it survives real conditions
Dashboards conventionally use colour to signal state — red for a problem, green for normal. Colour alone is insufficient for a meaningful proportion of users with colour vision deficiency, and it degrades further on projected screens, in bright rooms and on poorly calibrated monitors.
The fix is redundant encoding: colour supported by shape, position, an icon or a label, so the signal survives if colour is lost. This is an accessibility requirement and it also produces a more robust dashboard for everyone, since the meaning no longer depends on display conditions.
The related discipline is reserving alert colours for actual alerts. When the brand accent is also the warning colour, or when several statuses share a palette, the viewer has to read carefully to determine whether anything is wrong — which defeats the purpose of a display designed to be understood at a glance.
Questions, answered.
Dashboard UX design structures data screens around the decision they support — selecting metrics that would change that decision, directing attention to anomalies, and providing a path from signal to cause.
Usually because everything on it looks equally important, the numbers have no baseline to compare against, and noticing a problem offers no way to investigate it.
Fewer than it does. A useful rule is that adding a chart requires nominating one to remove, which forces the comparison that additive requests never do.
Practically, yes. A bare figure cannot be interpreted — nobody knows whether 847 is good without a baseline, a trend or a target alongside it.
Show much less. A phone can carry a few critical indicators and an alert; attempting the full desktop dashboard on a small screen produces something nobody reads.
As many as drive decisions for its intended user, which is usually far fewer than the number available. A monitoring view might need five or six. An analytical tool can support more because the user is actively exploring rather than scanning. The useful test is whether every visible metric would change someone’s behaviour if it moved; anything that would not belongs elsewhere.
Rarely. An executive needs outcomes and trends; an operational manager needs today’s queue and exceptions; a specialist needs detail in their own area. A single view serving all three either overwhelms the first or under-serves the third. Role-based defaults, with the ability to look further if wanted, generally work better than one shared screen.
Settle the definition, then display it where the figure appears. Most disputes about dashboard numbers are definition disputes rather than data errors — two teams counting slightly different things and each believing the dashboard is wrong. A visible definition at the point of use resolves the argument permanently and costs a tooltip.
Only where decisions are made in real time. Live updating on a dashboard reviewed weekly adds cost and distraction without benefit, and constantly shifting figures make trends harder to read. What every dashboard does need is a clear indication of how current the data is, since acting on figures assumed to be live but actually hours old is worse than knowing they are delayed.
Usually one of four things: it answers questions nobody asks, it takes too long to find the relevant figure, its numbers are not trusted, or it was built for a decision that has since changed. The trust problem is the hardest to recover from — once a user has been misled by a figure once, they verify elsewhere from then on and the dashboard has lost its purpose.
Still deciding if dashboard ux design is right for you?
Talk to UsAdding a Chart Costs Nothing
Someone asks for a metric on the dashboard. It exists in the warehouse, the query is simple, and adding a tile takes an afternoon. There is no reason to refuse.
What is not visible at that moment is the cost, because it lands somewhere else: on every future reader, who now has one more thing competing for the attention that should have gone to the number that mattered.
Twenty such decisions later, the dashboard contains everything and communicates nothing — and every individual addition was reasonable.
Free Dashboard Review
Find Out What Your Dashboard Is For
Send us your dashboard and tell us who uses it. We will identify which metrics support a decision and which are taking up attention.
Claim Your Free Marketing Audit
Enhance Your Brand Potential At No Cost!
- Expect a response within 24 hours
- NDA available upon request
- Dedicated product specialists