Metrics

Metrics blocks are the cards a dashboard is built from: a figure, the comparison that makes it mean something, and the chart underneath showing how it got there. Eight of them cover the shapes that carry most reporting, a KPI tile with its delta, rows lying down so long labels stay readable, two series ranked against each other, a week of stacked time, a ring of named shares, a year of trend bands, an hour-against-weekday matrix, and a zero line with a month's two sides either side of it.

The axis is where these go wrong quietly. Ticks spaced evenly on screen but not evenly in value will turn a flat series into a climb, and nobody reading the chart will know. The same care applies to the comparison: a figure with no stated period is a number, not a measurement. Say what the delta is measured against, keep the scale linear unless the label says otherwise, and let the reader reach the underlying rows when the shape raises a question a chart cannot answer.

18 blocks

Metric card

One headline figure with the months behind it: point at a month and it becomes the figure

Monthly rows

A short run of months you read down, like a table that draws

Ranked bars

Two series across a few categories: a legend that also switches them off

Day stack

A week of tracked time: what was booked and what was left

Donut breakdown

Where a total went: five parts at most, each named beside its share

Area trend

A running level over a year, with what is forecast set apart

Launch plan

A plan of tasks over weeks: what is done, what is late and where today falls

Hour heatmap

When in the week people show up, hour by hour

Diverging months

Money in against money out, month by month, with the balance

KPI card row

Four headline numbers over the same period, side by side

Figure strip

A few related figures read as one line, each against the period before

This week vs last week

A day-by-day comparison with the week before

Forecast with a target

Whether the year will land on its goal, with the likely range

Latency

A response time against its limit, with the traffic behind it

Noise level

A live reading on a fixed scale, from low to high

Score gauge

A score out of a total that is built from parts

Sign-up funnel

People moving through ordered steps, and who left at each

Service uptime

The health of a few services day by day, with the incidents behind it

01

Using Metrics

Reach for it when

  • The top row of a dashboard, where each tile answers a single question
  • A report someone scans before deciding whether to open the table underneath
  • Comparing two series across a few categories, with a legend that can switch one off
  • Looking for a pattern across two axes, such as hour against weekday

Reach for something else when

  • The reader needs the individual records: a chart shows shape and a table shows rows, so link one to the other rather than asking the chart to do both
  • There are only two or three values to compare; write them as a sentence, because a chart of three bars costs more attention than it returns
  • An operational signal that has to be acted on within seconds: a status page or an alert is the right surface, since a trend chart is read rather than watched
02

Where it shows up

01

Revenue dashboards

The tile a finance page leads with. One figure, a delta pill saying which way it moved, bars for the months behind it with an average rule across them, and month pills that move the selection without a reload.

02

Time tracking reports

A week of logged hours where the question is what was booked and what was left. Columns split hatched over solid, a legend carries each side's total, and an average rule shows which days sat under the usual load.

03

Staffing patterns

A matrix of hour against weekday, read as a shape rather than a series. Four tint steps are enough to show where a support queue fills; selecting a column or a cell reads the figure out for that shift.

04

Cash flow statements

Money in above a zero line and money out below it, month by month. The shared baseline is what makes the two sides comparable, and capping the unselected columns keeps the picked month legible without a separate highlight colour.

03

Questions

What makes a metric tile readable?

A comparison. A number on its own (revenue, signups, latency) cannot be judged, because the reader has nothing to judge it against. Pair the figure with a delta and name the period both belong to. If the comparison is too noisy at that window to mean anything, widen the window rather than dropping the comparison and leaving a bare number.

When is a donut better than a bar chart?

All blocks
The Orvain Dispatch

Premium components, free in your inbox.

Join the Orvain Dispatch and get production-ready components, free to keep: React and Vue code plus the Figma frame, yours to ship.

Just components, never spam. You can unsubscribe anytime.