Skip to content

Data & Analytics

Dashboarding: show what needs attention, and leave the rest out

Most dashboards fail by addition. Every chart anyone asked for, all the same size, until nothing stands out and nobody opens it. A useful one shows the few things that need a decision, whether you're on track, and how old the data is.
Discuss your systemSee one tile taken apart ↓
Needs attention · 2 jobs past their window · EastsideBooked revenue · 92% of month targettarget1 OcttodayData as of 06:10 · call tracking late

Forty charts, no decisions

Dashboards grow the way spreadsheets do. Someone asks for leads by day, someone else wants revenue by service, a third person wants a map. Each request is reasonable. A year later there are forty charts, all the same size, and the one number that needed attention this morning looks exactly like the thirty-nine that didn't.

Dashboarding is the editing. Who looks at this screen, how often, and what do they do differently because of it? Whatever doesn't change an action moves to a report or goes.

Four things worth a place on screen

  1. State

    Where things stand now

    Today's open jobs, this month's revenue, the queue. One number, with its target beside it.

  2. Trend

    Which way it's heading

    The line toward the target, not twelve months of history. Enough to tell “on track” from “drifting”.

  3. Exception

    What needs someone

    Jobs past their window, an unassigned lead, a location below plan. Exceptions get the strongest visual weight on the page, and a link to the records.

  4. Context

    Whether to believe it

    When the data last loaded, which source is late, what the metric means. A number without its freshness invites the wrong decision.

One tile, taken apart

Every element on a good tile answers a question the reader would otherwise have to ask.

  1. Value

    The number, large, in the unit people use. Minutes, not seconds.

  2. Comparison

    Against the target, not just yesterday, so the value means something.

  3. State

    On track, at risk, or off track, in words as well as colour.

  1. Trend

    Fourteen days with the target line. Enough to see drift, not enough to distract.

  2. Freshness

    When the data behind it last loaded.

  3. Next step

    A link to the records behind the number, so attention turns into action.

Same data, three people

One model underneath, three screens on top. Each person sees the horizon they work on and the exceptions they can act on.
Asks
Are we on track this month?
Looks
Weekly

Sees

Booked revenue

92% of target, trend toward month end

Margin by service

water heaters up, drains flat

Locations below plan

Riverside, with the reason

What we took off the screen

An operations dashboard that had grown to twenty-two charts. Nothing was lost; it moved to where it's used.
Leads by day, as a pieNobody acted on the daily splitThe monthly marketing report
All-time revenueNever changes a decision this weekThe quarterly board pack
Twelve-month trend for every metricHid this month's driftOne line, on the metric's own page
Map of every jobPretty, rarely openedThe dispatcher's view, filtered to today

What dashboarding hands off

In

Metrics
defined, with targets
Freshness
last load per source
Audiences
who looks, how often, why
Dashboarding
state · trend · exception · context

Out

Views
one per audience
Exceptions
linked to the records
Context
freshness beside each number

Worth building when someone checks the numbers every day

Good fit

  • Someone needs to know, daily or live, whether things are on track.
  • There are exceptions worth acting on quickly: late jobs, unassigned leads, a location behind plan.
  • The metrics underneath are defined, or are being defined.

Not the right fit

  • The numbers are reviewed monthly; a well-built report serves better.
  • The data underneath isn't trustworthy yet. A dashboard will display the problem more attractively.

Where dashboards do their work

Next to dashboarding

Questions about dashboarding

How many charts should a dashboard have?

As few as the person using it needs to act. For an operations screen that's often five to seven elements. If a chart hasn't changed a decision in a month, it probably belongs in a report.

Should every role get its own dashboard?

Usually each role gets its own view of the same model: the owner sees the month, a manager sees today's exceptions, a dispatcher sees the next hour. The numbers agree because they're the same numbers.

Can you rebuild the dashboards we already have?

Yes, and it usually starts with removing things. We look at who opens which screens and what they do next, keep what drives an action, and move the rest to reports.

What tool do you build dashboards in?

Often Metabase or the BI tool you already use; for operational screens inside your own software, a custom view. The design work is the same either way.