The idea
A dashboard is a page somebody opens to make a decision, to check that nothing is on fire, or to find out why a number moved. The first thing to know about any dashboard is which of those it is, because the answer decides everything else: which numbers sit at the top, how many filters it has, how often it refreshes, what each title says.
Think of the page as a short argument rather than a store window. It has a point, evidence that supports it, and detail for the reader who wants to check. If a chart cannot say what it adds to the argument, it comes off the page.
Figure 1
From a decision to a finished page
- Purpose What decision does this page serve?
- Audience Who opens it, how often, for how long?
- KPIs Three to six numbers they act on
- Layout Summary first, detail below
- Titles Each one states the insight
- Start with the decision. "Check that nothing is on fire", "decide where to spend next quarter" and "find out why returns jumped" are three different pages.
- The audience fixes the vocabulary and the rhythm: a CEO glances once a week, a store lead looks every morning, an analyst digs for an hour.
- Now pick the KPIs: the few numbers that person would change their behaviour for. A number nobody acts on is a decoration.
- Lay the KPIs out as a summary row, then the charts that explain them, then the detail table. The eye reads top-left to bottom-right.
- Last, write each title as the conclusion of its chart. The reader gets the point before they decode the picture.
Why it exists
Without that discipline, dashboards grow by request. Someone wants a chart, someone else wants a filter, and a year later there is a page with twenty visuals that nobody opens. Every chart is correct and the page is useless, because the reader has to work out the point themselves.
Harbor & Pine has three readers of the same sales data. The CEO wants to know if the business is healthy, once a week. A store lead wants to know what needs attention today. An analyst wants to dig. Three purposes, three pages, and no amount of clever design makes one page serve all three.
How it works under the hood
The pattern has four moves, and they apply whatever the tool.
- Name the decision. One sentence: "The CEO decides where to invest next quarter."
- Choose KPIs for that reader. Three to six measures they would act on. Each is a measure plus a comparison (last year, a target, a threshold), because a bare number has no meaning.
- Lay the page out from summary to detail. KPI row at the top-left, where the eye lands, the charts that explain them next, the detail below. Group related visuals and keep the filters in one place.
- Write titles as conclusions. A title that names the insight lets the reader get the point without decoding the chart.
The KPI step shows why the comparison matters.
Figure 2
Three KPIs for the executive page, 2025 against 2024
| KPI | Year 2025 | Year 2024 | Change |
|---|---|---|---|
| Sales | $2,300,169 | $1,993,604 | +15.4% |
| Orders | 4,295 | 3,673 | +16.9% |
| Average order value | $535.55 | $542.77 | -1.3% |
- A KPI is a measure plus a comparison. Alone, $2,300,169 cannot tell the reader whether to worry. Set beside last year, it can.
- Sales are up 15.4%. That is the headline, and a CEO would stop reading here.
- The other two KPIs explain it: there were 16.9% more orders, while the average order slipped by 1.3%.
- The insight is in the pair: growth came from more customers ordering, not from bigger baskets. Neither number says that alone.
Three kinds of page
The decision gives the page a type.
- Strategic: slow-moving numbers against long-term goals, for executives. Yearly and quarterly trends, few KPIs, little interaction.
- Operational: what is happening now, for the people running it. Current values against thresholds, exceptions highlighted, frequent refresh.
- Analytical: open questions, for analysts. Many filters, drill-downs and detail tables, because the reader brings their own question.
Mixing them is the usual mistake. An executive page with eight slicers is an analytical page that forgot its audience.
Layout is reading order
People scan a page in a Z: top-left, across, then down. Put the summary where the scan starts and the detail where it ends. Leave space between groups, because empty space tells the reader which visuals belong together, and line everything up on a grid, since a crooked edge reads as carelessness before anyone looks at a number. Keep one consistent colour for one meaning across the whole report: if gold means 2025 on the first page, it should not mean Furniture on the second.
Before you publish
Read the page as the person it is for, with the page closed to everyone else's questions. Can they tell in five seconds what it is about? Does every KPI have a comparison? Does every title make a claim that the chart supports? If the answer to any is no, change the page, not the reader.
Report, dashboard or analysis
In Power BI the words have precise meanings. A report is a multi-page file made in Desktop that the reader can filter and explore. A dashboard (the Service feature) is a single canvas of tiles pinned from reports, not interactive in the same way, built to be glanced at. An analysis is a piece of work done once to answer a question, often a report that was never meant to be kept. Most of what people call "the dashboard" is a report page.
Titles that carry the story
A good story has four parts: the context (what we are looking at), the conflict (what changed or is wrong), the insight (why), and the recommendation (what to do). A title can carry the insight, and the subtitle the recommendation.
Figure 3
Completed sales by year
- Title A, "Sales by year", describes the chart. It is true and tells the reader nothing they cannot see.
- Title B, "Sales grew 15% in 2025, after 25% in 2024", states the conclusion. The chart now proves a sentence.
- The sentence carries a promise: the bars must show it. 2024 was 25% above 2023 and 2025 15% above 2024, and the columns are on a zero baseline so the heights are honest.
In Power BI Desktop
Build the KPI row from three measures in Home → New measure, and for a comparison use
CALCULATE([Sales], DATEADD('Date'[Date], -1, YEAR)) or a prior-year measure. Titles are set in
Format → Visual → General → Title, and a title can itself be a measure so it updates with the
selection. Use View → Gridlines and snap to line the visuals up.
Common misconceptions
"A good dashboard shows everything"
A dashboard that shows everything shows nothing. Each extra chart competes with the one that matters for attention. The test is subtractive: for each visual, ask what decision it supports, and remove it if the answer is "none".
"Titles just label the chart"
A label describes. A conclusion tells. "Sales by year" forces the reader to look at three bars and work out what you already know. "Sales grew 15% in 2025" tells them what to see and lets them check it against the picture. Write titles you could defend in a meeting.
"More KPIs make the page more useful"
Past six or so, the eye stops seeing any of them. The CEO's page needs the three numbers that would change a decision, with the comparison that gives each one meaning. Everything else goes on a second page, one click away.
Check yourself
One question. Answering it ticks this lesson off on the map; it never locks a level.
Right. That is an operational page: current numbers, a standard to compare with, and the exceptions pulled forward, because the decision is whether to act now.
Not quite. Trends belong on a strategic page. At 8 a.m. the question is what is wrong today, and a three-year line does not answer it.Try another answer.
Not quite. That is an analytical page, for someone with an hour and a question. It makes a busy person do the work of finding the exception.Try another answer.
Not quite. A yearly figure changes too slowly to prompt any action this morning, and it has no comparison to say whether it is good.Try another answer.
Lesson done: it is ticked on the map.