KPI Dashboard Template: Every Tile Needs a Comparison and an Owner

Thirty-eight tiles, eleven red, no decisions

Picture a monthly business review. The dashboard on the screen has thirty-eight tiles. Eleven are red. The CEO points at one of them: service revenue, red. “Is this bad?” The analyst who built it explains that red means 2% below budget. Was it red last month? No, green. Is 2% unusual for this line? Nobody knows. Twenty minutes later the meeting moves on, and nothing has been decided.

That dashboard had plenty of data. What it lacked was two things per tile: a comparison that tells you whether the number is news, and a name that tells you who does something about it.

A KPI dashboard is a single screen of the few measures a team needs to steer toward its objectives, each shown against a comparison (a target, a trend, a normal range) and each owned by a person who acts when it moves. Stephen Few's definition is still the best one: “a visual display of the most important information needed to achieve one or more objectives; consolidated and arranged on a single screen so the information can be monitored at a glance” ¹.

The thesis of this page: a dashboard is the agenda of a decision meeting, drawn as pictures. A tile earns its place only if it answers “compared to what?” and names the person who acts when it moves. Everything else belongs in a report.

A number without a comparison is trivia. A red tile without an owner is decoration.

What a dashboard can do, and what it cannot

Few is precise about the limit. A dashboard's high-level summaries tell you “what's happening, but not why it's happening, just like the gauges, meters, and indicator lights on a car” ¹. Diagnosis takes further investigation ¹. So a dashboard is a trigger for a conversation, not a substitute for one. If your review meeting expects the screen to explain a miss, it will get a slide of excuses instead.

Two other limits follow from the single screen. First, choosing the measures is a separate job, done before anyone opens a design tool; our page on KPI examples that track strategy covers how to pick them and how to pair each with a counter-measure. Second, a screen holds only a handful of things a human can take in at once. Few points out that one of the great benefits of a dashboard is “the simultaneity of vision, the ability to see everything you need at once” ². Scrolling and tabbed views break it.

For the textbook definitions, see key performance indicators and leading vs. lagging indicators. This page is about the screen.

Two panels: what must change, and what must not break

Most dashboards mix two kinds of measure, and the mix is why nothing on them gets a decision. Some numbers track the few things the strategy says must change this year. Others track the health of the business you already run. They need different comparisons and different conversations.

Hewlett-Packard's planning materials, as Barry Witcher describes them, kept the two apart: hoshin plans for a very few breakthrough objectives, and a separate business fundamentals table with more numerous KPIs for incremental improvement ⁴. The same split works on one screen. Panel A holds the breakthrough measures, judged against a path to a target. Panel B holds the fundamentals, judged against their normal range. Our page on OKRs vs. KPIs explains why the two kinds of number behave so differently.

• Panel A: strategic bets (three to six tiles). One tile per measure the strategy needs to move, each with its target path. Question it answers: are we changing what we said we would change?
• Panel B: fundamentals (six to ten tiles). Revenue, margin, cash, safety, quality, service. Question it answers: is anything breaking while we change?
• Panel C: exceptions (zero to three lines). Anything outside its normal range this period, with the owner's one-line note. Empty is a good month.

The template

Start with a header block, then specify every tile with the nine fields below before you draw anything. A tile you cannot specify is not ready for the screen.

• Header: dashboard name; the decision meeting it serves; audience; refresh date; the person who owns the definitions.
• Footer: data sources and the date each was last refreshed.

Fields 4, 5 and 6 are where most dashboards fail. Fields 7 and 8 are where most dashboards stop being useful even when they are accurate.

Rules for the status color

Color is a verdict, so it needs a rule. Few's test for every number is blunt: “Compared to what? Is this good or bad? How good or bad? Are we on track?” ². A tile that shows quarter-to-date sales with no target and no history answers none of those questions ².

The common fix, red-amber-green against budget, creates a new problem. Every measure wobbles. NHS England's guidance on statistical process control makes the point when it explains its own status icons: for a measure that will sometimes meet its target and sometimes miss it by random variation, “in a RAG report this indicator would flip between red and green” ³. The same guidance warns against narrative that “describes insignificant changes from one data point to the next” ³. A tile that flips color every month on noise trains the room to ignore color, and then a real signal arrives red and gets the same shrug.

• Plot the history, not just the point. Twelve or more periods show the normal range at a glance. Last month versus this month shows almost nothing.
• Color on a written rule. For fundamentals: outside the normal range, or a sustained run in one direction. For strategic bets: off the target path by more than an agreed margin for two periods in a row.
• Show the variance directly. If the point is the gap to budget, show −10%, not two dollar figures the viewer has to subtract. Few calls the alternative expressing measures indirectly ².
• Round. $3.8M, not $3,848,305.93; excessive precision slows the viewer without benefit ².
• Never rely on red and green alone. Few notes that red-yellow-green coding excludes color-blind viewers ². Add a symbol or a word.

A filled-in example

The company below is invented. Its strategy: grow recurring maintenance plans so that revenue depends less on emergency repairs.

Notice what the screen does to the meeting. The CEO's question changes from “is this bad?” to “do we pause the rollout?” Attach rate at 18% against a June path of 20% is a conversation about pricing, not about whether the number is worrying. And the exceptions line arrives with a decision attached, which is the whole point of having it.

Build it in two weeks

1. Name the meeting firstWrite down which recurring meeting the dashboard serves and which decisions that meeting makes. A dashboard with no meeting has no audience.
2. List the choices, then the measuresFor Panel A, one measure per strategic choice. For Panel B, the fundamentals that would sink the business if they broke. Stop at about fifteen tiles in total.
3. Specify every tile on paperFill in the nine fields for each tile. Any tile with no owner or no trigger goes back to the list or into a report.
4. Pull 12 or more periods of historyUse it to set each normal range. If the history does not exist, say so on the tile and start collecting it.
5. Write the color rules and get them signedThe executive who chairs the meeting agrees the rules before the first review, so color stops being a debate.
6. Run one review and pruneAfter the first meeting, remove every tile nobody discussed and nobody would miss. Add nothing for a quarter.

The case for the big dashboard, and the politics of a tile

The strongest objection to a fifteen-tile screen is that executives get blindsided by the measure that was left off. Better to show everything, with drill-downs, than to miss the problem in the forty-first metric. Modern tools make that cheap.

The objection is right that the missing measure is a real risk, and wrong about where to put it. Few's point about the single screen is that seeing everything together is what lets you spot connections, and that fragmenting the view sacrifices exactly that ². A screen of forty tiles is not a bigger dashboard. It is a report that nobody will look at long enough to see the connection. Handle the forty-first metric with the exceptions panel: every fundamental lives in a report and surfaces on the screen only when it leaves its normal range.

The harder problem at scale is political. A tile on the CEO's screen is status. Removing one demotes a team, and every function lobbies to get its number on. Three defenses work. One person owns the definitions, so no unit can quietly redefine its own tile. Tiles are reviewed once a year against the strategy, with a written reason for each that stays. And the owner of a tile presents it, which makes a tile a duty as well as an honor, and thins the lobbying fast.

One more second-order effect: a green-heavy screen is not good news by itself. If Panel A is green every month, either the strategy is working or the targets were set where they could not be missed. The quarterly strategy review is the place to ask which.

Before the first review

• The dashboard names the meeting it serves.
• Panel A has one tile per strategic choice and no more than six.
• Every tile shows a comparison: a target path, a normal range, or both.
• Every status color comes from a written rule.
• Every tile has one owner and a trigger with an action.
• Every tile tied to pay has a counter-measure beside it.
• Nothing needs scrolling.

Monday morning: open your current dashboard and, for each tile, write the owner's name and what happens if it turns red. Every tile where you cannot write both is a candidate to leave the screen.

The test of a dashboard is not what it shows. It is what the meeting does in the five minutes after it goes up. If the answer is “discusses whether the numbers are bad,” the dashboard did not do its job. If the answer is “decides,” it did.

Questions

What should a KPI dashboard include?

A header naming the meeting it serves and its audience, then two panels: three to six measures the strategy needs to change, each against a target path, and six to ten fundamentals, each against its normal range. Every tile needs a definition, a comparison, a written status rule, one owner and a trigger that says what happens when the rule fires.

How many KPIs should be on a dashboard?

Few enough to fit on one screen without scrolling, because seeing the measures together is what lets you spot connections between them. In practice that is about fifteen tiles: up to six strategic measures and six to ten fundamentals. Everything else belongs in reports, surfacing on the dashboard only when it moves outside its normal range.

What is wrong with red-amber-green status?

Nothing, if the color comes from a written rule against a normal range. The problem is coloring by variance to budget alone: every measure varies, so a tile can flip between red and green on random noise. The room learns to ignore color, and then misses the real signal. Show the history and color only when a point leaves its usual band.

What is the difference between a dashboard and a report?

A dashboard is a single screen of summaries, arranged so you can monitor the few things that matter at a glance and decide whether to act. It tells you what is happening, not why. A report holds the detail you need to diagnose a problem once the dashboard has flagged it. Most failed dashboards are reports squeezed onto a screen.

Where to go next

• KPI Examples That Track Strategy, Not Activity
• OKRs vs. KPIs vs. Goals
• Quarterly Strategy Review: Agenda and Template
• Leading-Indicator Dashboard
• Balanced Scorecard: Framework & How to Apply It
• KPI vs. KRI