Choose the visual by the question
Every dashboard decision starts with what the viewer must answer. Comparing values across items calls for a bar or column chart. Showing a trend calls for a line. Displaying a part of a whole calls for a stacked bar or a donut. Reading precise figures calls for a matrix or a table. Match the visual to the job and the dashboard stops needing a footnote.
The mistake is reversing the order: pick a pretty visual, then hunt for data to fill it. That habit produces dashboards where the viewer cannot tell what changed or which category is worst. Start from the question, then choose the chart that answers it fastest.
Comparison visuals: bar and column
Add a bar chart, put Total Sales = SUM(Sales[Amount]) in Values and Region in the Axis. West, East, North, and South appear as bars you can rank at a glance.
Click the visual, choose Sort by Total Sales. East jumps to the top at 402,110 so the viewer sees the leader without scanning.
When categories are quarters or ages, a column chart reads left to right naturally. Keep bars for pure category ranking to avoid confusing the two.
I recommend limiting a bar chart to about ten categories. Past that the bars get thin and the viewer cannot compare. Group small items into an Other bucket before charting.
Trend visuals: line and area
Add a line chart with Month from the Calendar table on the Axis and Total Sales in Values. The November peak of 142,300 shows as a clear rise.
Drop Category into Legend to overlay three lines. Now the trend answers both 'when' and 'which category' in one visual.
An area chart works for one or two series. With five categories the fills overlap and hide the lines, so switch back to a plain line chart.
Part-of-a-whole visuals
A donut of Sales by Category shows Electronics at 41%, Office at 31%, Home at 28%. Use it only when the shares matter more than the exact totals.
When you also need to compare totals across regions, a stacked bar of Category within Region beats a donut because lengths stay comparable.
A common mistake is using a pie chart with seven slices. Humans cannot compare angles, so the viewer guesses. A bar chart of the same shares is read accurately in a glance.
When exact numbers matter: matrix and table
Drop Region on Rows, Category on Columns, and Total Sales in Values. The grid shows 318,940 for West Electronics and lets an analyst pull the precise figure.
On the matrix, set data bars on the Values field. High cells fill darker, so outliers like the November spike stand out without reading every number.
A raw table is best when the viewer will export to Excel. On a dashboard meant for reading, a matrix with bars communicates faster.
Mistakes that sink a dashboard
A common mistake is putting three KPIs, a map, and a scatter on one page 'just in case'. The viewer leaves confused. Pick three visuals that answer three linked questions.
This will break if you use a 3D pie or shadowed bars to impress. Perspective distorts lengths and the viewer misreads the magnitude of every value.
Assign one color per series, not a rainbow per bar. Random colors make the eye hunt and hide the real signal in the Sales by Region chart.
Do not put Category on a map just because the visual exists. A map needs real geography like Region or Country; otherwise use a bar chart.
Worked example: a one-screen sales dashboard
For a Sales model with 42,000 rows, Region, Category, Month, and Amount, build three visuals: a bar chart of Total Sales by Region (East leads at 402,110), a line chart of Total Sales by Month (November peaks at 142,300), and a matrix of Region by Category with data bars. The viewer reads the leader, the trend, and the detail without a single table of raw numbers. Total across all visuals reconciles to 1,245,300.
If you know Excel charts, here is the shift
Excel chart selection follows the same logic, but Power BI adds slicers and cross-filtering so one click on a bar filters the whole page. Our Excel chart selection guide covers the static side, and our dashboard guide in Excel shows how to lay out a similar one-screen report before you move it to Power BI.
The principles transfer: answer the question, limit categories, skip 3D. Power BI just makes the chosen visual interactive across every other visual on the page.
Scatter, map, and when to skip them
A scatter chart shows the relationship between two measures, such as DiscountPct on one axis and Total Sales on the other, with each bubble a product. Use it to spot clusters, like high-discount items that still sell strongly. It answers a correlation question that no bar chart can, but only when you truly have two numeric axes to compare.
A map visual needs real geography. If your data has a Region or Country column, a filled map shades West darker as its 318,940 total grows. Without a geographic field, a map is decorative noise, so switch to a bar chart. The rule: maps communicate only when location is the actual dimension being analyzed.
Put Total Sales on the Y axis, DiscountPct on the X axis, and ProductName in Details. Each point is one product, and outliers reveal themselves without a single filter.
Drag Region to the Location field and Total Sales to the Color field. West shades deepest because its 318,940 leads the regions in the sample model.
I recommend reserving the scatter for audiences comfortable reading two axes at once; for executives, a ranked bar chart of the same totals communicates the point in a glance without a legend explanation.
Layout and visual hierarchy
Selection is half the battle; placement is the other half. Put the most important visual, usually the headline KPI card, top left where the eye lands first, then support it with the trend and the detail. A dashboard that buries the key number at the bottom forces the viewer to hunt and undermines the whole design.
Use a consistent visual hierarchy: one large visual, two medium, and a grid of small ones, rather than six equal tiles. Alignment to a grid makes the page feel intentional, and whitespace between cards prevents the viewer from merging two unrelated numbers.
Drop a card with Total Sales = 1,245,300 in the top-left corner so the headline answers the first question before the viewer scrolls.
Snap the bar chart and line chart to the same width and a shared left edge. Aligned axes let the viewer compare Region and Month without moving their eyes far.
Accessibility and color choices
Color is data, not decoration. Use a single highlight color for the value the viewer should notice, such as East in the bar chart, and neutral grays for the rest. This directs attention instead of scattering it. Avoid red-green pairs, which exclude color-blind viewers, and avoid gradients that imply a precision the data does not have.
Also add alt text to each visual so screen readers can describe it. Power BI lets you write a description per visual; a one-line note like 'Bar chart of Total Sales by Region, East highest at 402,110' makes the dashboard usable for everyone and documents the intent for future editors.
In the bar chart formatting, set all bars to gray and only East to blue. The viewer's eye lands on the leader without reading every label.
Select each visual, open the Alt Text pane, and write one sentence describing the chart and its key takeaway so the report meets accessibility standards.
Choosing visuals for mobile and print
A visual that shines on a 27-inch monitor can fail on a phone. On mobile, favor cards and simple bars that survive a narrow width, and drop the matrix with twenty columns that becomes unreadable. Power BI's mobile layout lets you place a few key visuals in a single column so the viewer gets the headline without zooming.
Keep a Total Sales card, a Region bar chart, and a Month line chart for the phone layout; they read clearly at 400 pixels wide without horizontal scrolling.
Exclude the full Region-by-Category matrix from mobile, or replace it with a filtered top-five bar so the small screen shows the signal, not the noise.
For print or PDF export, avoid visuals that rely on interactivity, because a printed page cannot be clicked. A static bar chart with data labels communicates the comparison better than a slicer-driven visual that arrives at one filtered state nobody chose.
Myths that lead to bad dashboards
One lasting myth is that more visuals mean a more complete dashboard. In practice, a page with eight charts trains the eye to ignore all of them. Three well-chosen visuals that answer three questions beat eight that answer none, because attention is the scarce resource, not screen space.
Before adding a visual, write the question it answers. If you cannot state one, cut it. The Sales dashboard needs only leader, trend, and detail to be useful.
Remove gauges and spinning icons that show a single percentage; a card with the number and a tiny delta does the same job with less clutter.
I recommend showing a draft to one real viewer and watching where their eyes go first. If they miss the key number, the layout is wrong, not their attention, and you should reorder the visuals.
Choosing visuals with stakeholders
The best chart decision often comes from asking the stakeholder one question: what will you do with this number? If they will decide whether to act, a clear bar chart with the leader highlighted is enough. If they will present to executives, a single KPI card plus a trend line communicates the story without detail they will not read. Matching the visual to the decision prevents the classic dashboard that impresses and informs no one.
Before building, ask what action follows the chart. The answer tells you whether you need a comparison, a trend, or a single headline, which selects the visual directly.
Build one visual and show it within a day. A stakeholder correcting 'I needed it by region, not by month' early saves a full rebuild of the page later.
I recommend writing the one-sentence takeaway of each visual before you pick the chart; if you cannot state it, the visual has no job and should be cut from the dashboard.
Avoiding visual debt over time
Dashboards grow by addition and rarely by subtraction, which is how a clean three-visual page becomes a cluttered ten-visual one. Set a rule that adding a visual requires removing or justifying another, and review the page quarterly. The goal is a report whose every element earns its place, not a museum of every chart type Power BI offers.
Every three months, open the dashboard and ask which visual nobody clicks. Move unused ones to a hidden detail page or delete them to keep the main view focused.
Save an approved layout with the KPI card, bar, and line as the standard start, so new pages stay consistent and the team does not re-litigate chart choice each time.
Key takeaways
Chart selection is a question of intent, not decoration: pick the visual that answers the viewer's question in a glance, limit the page to the few that matter, and verify the layout with a real reader before you ship. A bar for comparison, a line for trend, and a matrix for exact figures cover most dashboard needs.
Consistency and restraint beat variety. Align to a grid, use one highlight color, and cut any visual without a stated job, and your dashboard will inform instead of impress. The measure and model behind the charts matter just as much, so build those first and let the visuals do the simple job of showing the answer.
Next steps
Sketch the three questions your dashboard must answer, then assign one visual to each. Build the model first so the measures have clean fields to read, then add slicers last so viewers can pivot the view. The data modeling guide sets up the Sales and Calendar tables these charts rely on, and the DAX measures guide gives you the Total Sales pattern used in every example here.
If you are comparing tools, the Power BI vs Tableau guide shows how visual selection works in both, and the PL-300 guide lists dashboard design as a scored skill.
