Dashboard KPI Design
Structure pages the way decisions are made
A KPI dashboard is a decision page, not a data gallery. Hierarchy: north-star on top, drivers beneath, comparators beside. Clutter is the silent killer of decision speed.
What you will learn
- Structure a KPI page with hierarchy
- Choose comparators (MoM, target, forecast)
- Reduce clutter without losing context
What is this?
KPI design arranges metrics by decision priority: one north-star card at the top, the driver cards beneath it, and context (target, previous period, sparklines) beside each. Comparators turn a bare number into a direction ('+12% vs last month', 'under target'). Every element must survive the 'so what' test.
Why do you need it?
- Executives scan; a wall of 30 numbers gets skimmed for the one that matters.
- Without comparators a number has no meaning — 98% is good or terrible depending on the benchmark.
- Hierarchy leads the eye: north star first, drivers after, context last.
- Decision-grade dashboards change actions, which is the entire point of analytics.
Mental model
A cockpit instrument panel
A KPI dashboard
Pilots get one primary attitude indicator, then the secondary dials, then the ambient details — never 20 equal gauges. Crucial instruments have a comparator (target line) baked into the display, like your speed tape versus the speed bug.
How it works
- Define the north star: the one number that gates the strategy (MRR, active usage).
- Lay driver KPIs below it that move the north star (activation, retention share).
- Give every KPI a comparator: prior period, target, or forecast.
- Add sparklines for trend and tooltips for drill detail.
- Strip non-decision decoration; every element answers a question.
Hierarchy of a decision dashboard.
Examples
A KPI card pattern
[MRR (USD)] [ +6.2% MoM ] [== plan line ==]sparkline of 13 months below, tooltip: defined in registry v4The card carries the number, the comparator (MoM %) and the plan line sparkline, with the definition one hover away — the dictionary semantics without the clutter.
In production
- An exec page leads with MRR and churn cards, each with a MoM % and a target sparkline; drivers (activation, expansion) sit below.
- A customer-success team's page starts with a gross-retention card and a churn-warning threshold band, drivers beneath.
- A marketplace board leads with GMV north-star, then take-rate drivers, each card with an actual-vs-budget comparator.
Best practices
- One north star per page; if there are two, split the page.
- Add a comparator to every card — a number without comparison is noise.
- Use hierarchy in layout: top-to-bottom = strategy to operation.
- Show trends in sparklines instead of giant data labels.
Common mistakes
- Forty cards equals zero priorities, whichever exec gets the remote reads none.
- Percentages without denominator or comparison.
- Decorating with visual noise that costs seconds per glance.
- Live-ish data in an executive context that causes thrash instead of decisions.
Common issues
Execs still ask for a separate reportThe page doesn't answer the decision the exec actually makes
Interview the consumer, re-map hierarchy and comparators to their decisions.
Sparkline trends disagree with cardsCard value is a different window than the sparkline
Bind both to the same time range and grain.
Users overload the page with filtersNo hierarchy of interactions
Keep three decisions per page and silo filters per card group.
Debugging tips
- Test the 'ten-second read': can a newcomer state the north-star and its trend?
- Check every comparator has a definition (previous period = which window?).
- Remove one element at a time and confirm the decision still forms.
- Log what actions the page actually drove; a dead metric is a defect.
Interview checkpoint
Quick quiz
Where does the north-star KPI belong in the layout hierarchy?
A comparator (last month, target) is optional decoration for KPI cards.