Skip to content
Tutorials Center
My learning

Topics

Lesson 35 of 56 ยท Visuals and Reports

Choosing the Right Visual

Match a business question to the right Power BI visual: bar, line, scatter, table, matrix, map, treemap and waterfall, with matplotlib and pandas stand-ins you can run.

Power BI ships with dozens of visuals, and it is easy to pick one because it looks impressive. A better habit is to start with the question the reader is asking and choose the visual that answers it with the least effort. "Which category sells most?" is a comparison. "Are we growing?" is a trend. "Does a higher price lower the volume?" is a relationship. Each type of question has a visual that fits it, and a handful of visuals covers nearly every report you will build.

This lesson goes through the core visuals, what each is good at and where it misleads. You build visuals by dragging fields onto the canvas in Power BI Desktop, so the steps are described in text. Power BI cannot run on this site, so the Python examples are stand-ins: matplotlib draws the same chart types from the same kind of data, and pandas builds the matrix. If you have not built a first chart yet, start with Build Your First Report in 10 Minutes.

Start with the question

Compare Which is bigger? Use bars.
Trend How does it change? Use a line.
Relationship Do two numbers move together? Scatter.
Detail Need exact values? Table or matrix.

The question picks the visual, not the other way around.

1 / 2

Bar and column charts: comparing categories

A bar chart (horizontal) or column chart (vertical) shows one value per category. The eye compares lengths very accurately, which makes bars the safest choice for ranking and comparison. In Power BI you choose Clustered bar or Clustered column, drag a category field to Y-axis (or X-axis) and a measure to X-axis (or Y-axis).

Some rules that make bars work:

  • Start the value axis at zero. A truncated axis exaggerates differences. Bars encode value by length, so a cut-off bar lies.
  • Sort by value, unless the categories have a natural order such as months or sizes. Use the ... menu on the visual and choose Sort axis.
  • Use horizontal bars for long labels such as product names; vertical columns suit short labels and time.
  • Stacked bars show parts of a whole, but only the bottom segment is easy to compare. Use them for one or two series, and prefer clustered bars when exact comparison matters.

Pie and donut charts also show parts of a whole, but people judge angles poorly. They can work for two or three slices that differ a lot. For anything more, a sorted bar is clearer.

A line chart connects points in order, so it shows direction, speed and turning points. Put a date field on the X-axis and a measure on the Y-axis. Use a date hierarchy (see Date Tables and Time Intelligence Setup) so readers can drill from year to quarter to month.

  • Use a line when the x-axis has an order, nearly always time. Do not connect unordered categories.
  • With more than four or five lines, the chart turns into spaghetti. Use small multiples or highlight one series.
  • A line's axis does not have to start at zero, but be honest about it, and consider adding a reference line such as a target or average from the Analytics pane.
  • An area chart is a line with the area filled in. Use it for volume, and stacked areas for composition over time.

Scatter plots: relationships between numbers

A scatter chart plots each item as a point, with one measure on the X-axis and another on the Y-axis. Put the item (product, customer, store) in Values. Size and colour can add a third and fourth variable (a bubble chart), and a Play axis can animate over time. Scatter charts answer "do these move together?" and reveal outliers: the product that sells a lot at low profit, the store that is off the pattern.

Remember that a visible relationship is not proof of cause. Two measures can move together because a third factor drives both.

Table and matrix: exact values

Sometimes the reader wants numbers, not shapes. A table shows rows and columns of fields with the values to the right. A matrix is a pivot table: row fields down the side, column fields across the top, values in the cells, with subtotals and grand totals. The matrix supports drilling into the row hierarchy, and conditional formatting can colour cells as a heat map (covered in Conditional Formatting, KPIs and Cards).

Use a table to list items with several attributes, and a matrix to compare two dimensions, such as category by year. Do not build a wall of numbers when a chart would say it faster. A table next to a chart works well: the chart gives the pattern and the table gives the exact figures.

Maps: where things are

Power BI offers a map (bubbles on a geographic background), a filled map (regions shaded by value) and an Azure map with more options. Maps require geographic fields: assign a Data category such as City, Country or Latitude/Longitude in the Column tools ribbon so locations resolve correctly. Use a map only when location is the point. If the real question is "which region sells most?", a sorted bar chart is easier to read, because the size of a region on a map has nothing to do with its sales. Bubble maps can also be unreliable when place names are ambiguous, so prefer coordinates or complete addresses when you have them. Which map visuals your organisation can use may depend on tenant settings.

Treemap and waterfall

A treemap fills a rectangle with tiles whose area is proportional to a value, grouped by category. It suits many parts of a whole, such as revenue by product across hundreds of products, where a pie would fail. It is poor for precise comparison, since areas are harder to judge than lengths. Use it for overview, then let readers drill or cross-filter into the detail.

A waterfall chart shows how a starting value turns into an ending value through a series of increases and decreases: profit bridges, headcount changes, budget to actual. Put the steps in Category and the amounts in Y-axis; increases and decreases are coloured differently, and a total bar closes the bridge. A finance report uses it constantly, and you will build one in Project: Profit and Loss Report.

Seeing the three basics side by side

The first stand-in draws a bar chart, a line chart and a scatter chart from small data sets, then prints the one number each chart is meant to reveal. It uses matplotlib; on this site the figure appears as an image under the code.

Python Compare, trend, relationship

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The bar chart is sorted, so the biggest category, bikes with 48% of revenue, leads. The line chart shows growth of 65% from January to June, and the scatter shows that higher prices go with fewer units. Notice that each chart answers its question without a legend or extra explanation. That is the test of a good choice.

A matrix as a pivot

The matrix visual is a pivot table, so pandas' pivot_table is a faithful stand-in. Rows are categories, columns are years, and margins=True adds the subtotals Power BI shows by default.

Python A matrix with totals

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The region column is not in the matrix, so its rows are summed into each cell, just as a matrix visual aggregates fields you do not place. Adding Region under Category would be the equivalent of a second level in the matrix row hierarchy.

A waterfall bridge

The waterfall visual does not need a separate calculation: it takes the steps and draws the running total. The stand-in builds the same chart in matplotlib by hiding the bar bottoms.

Python A waterfall from start to end

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Starting at 1,000, adding 300 and 80 and subtracting 120 and 50 gives an end value of 1,210. Green steps lift the total and red steps lower it. Each floating bar starts where the previous one ended, which is what makes the "bridge" readable.

More chart types: stacked, area, combo, bubble, donut, funnel, gauge and ribbon

The visuals above cover most questions. The ones below each solve a narrower problem. Know what they are for, and also when a plain bar or line chart would do better.

Visual Use it to show Watch out for
Stacked column / bar a total split into parts (sales by month, split by category) Only the bottom segment starts at zero, so the upper segments are hard to compare.
100% stacked column / bar how the mix changes (share of each category per month) Totals disappear; show them in a tooltip or a second visual.
Area chart the volume under a trend over time Use a stacked area only with few series; the order of the layers affects readability.
Line and clustered column (combo) two measures with different units on one time axis (revenue as columns, margin % as a line) The secondary axis can imply a relationship that is not there; label both axes.
Scatter with size (bubble chart) three numbers at once: X, Y and bubble size, with a color for the category More than a few dozen bubbles overlap; use a play axis for time.
Donut chart the share of a few parts of one whole More than five slices become unreadable; a bar chart is almost always clearer.
Funnel stages where a count drops from step to step (visits, sign-ups, purchases) Order the stages as the process runs, not by size.
Gauge one value against a minimum, a maximum and a target Takes a lot of space for one number; a KPI or card often says it better.
Ribbon chart rank changes over time: the ribbons connect categories from one period to the next, with the biggest on top Good for "who is first this month", poor for exact values.
Filled map values by region (country, state, postal code) with color intensity Large areas dominate small ones; normalize by population or use a bubble map.

To change a visual's type, select it and click another icon in the Visualizations pane: the fields stay in place whenever the new type has matching wells. Format options such as stacking, axis ranges and data labels live in the Format tab.

Formatting a matrix

A matrix is a table with a hierarchy on rows and columns, as in the pivot example above. In Format > Style presets choose a layout, then adjust: turn Row subtotals and Column subtotals on or off, set Stepped layout so child rows are indented, use Expand/collapse (the + icons) to drill into the hierarchy, and add Conditional formatting to the values (data bars, background colors). For readability, show only the subtotals you need and keep the row headers short.

Try it: the numbers behind three of these charts

The stand-in below computes what a 100% stacked chart, a funnel and a ribbon chart would display, so you can see why each is the right choice for its question.

Python Stand-in: 100% stacked shares, funnel conversion and ribbon ranks

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The 100% table shows that Helmets grew from 25% to 50% of the mix, even though their sales rose by only 250. The funnel shows where most people drop out (visits to sign-ups loses 88 percent). The ranks show Kids climbing from third to first, the kind of overtaking a ribbon chart makes visible with a crossing band.

Every chart type is a trade-off. When in doubt, start with a bar or a line, and switch only when the question needs the extra encoding.

Quick rules of thumb

  • One visual, one message. Give each visual a title that states the point ("Revenue grew 65% since January") and not only the field names.
  • Prefer position and length (bars, lines, dots) over area, angle and colour for accurate comparison.
  • Cut clutter: gridlines, borders, 3-D effects and too many colours distract. The next lesson, Formatting, Themes and Report Design, shows how to tidy a report.
  • If unsure, use a bar chart. If it is a trend over time, use a line.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. A manager asks: which of our 12 product categories sold most last year? Which visual fits best?

    Ranking categories is a comparison, and the eye judges bar lengths accurately. A pie with 12 slices is hard to read.
  2. 2. Why should a bar chart's value axis start at zero?

    A bar's length is the data. Truncating the axis makes small differences look large, so it misleads.
  3. 3. Which visual shows how a starting value changes into an ending value through increases and decreases?

    A waterfall chart bridges start and end with floating bars for each step, coloured by direction.
  4. 4. When is a map a good choice?

    A map shows where things are. To rank regions, a sorted bar is clearer, because a region's size on the map has nothing to do with its value.
  5. 5. What is a matrix visual closest to?

    A matrix puts fields on rows and columns, aggregates values in the cells and can show subtotals and grand totals, like a pivot table.

Summary

  • Start with the question: compare (bar), trend (line), relationship (scatter), exact values (table or matrix), location (map), parts of many (treemap), bridge (waterfall).
  • Bars should start at zero and usually be sorted; avoid pies with many slices.
  • Lines need an ordered axis, normally time, and few series.
  • Maps only help when location is the point; waterfalls explain changes from start to end.
  • Matplotlib and pandas stand-ins show the same chart types, but you build the real visuals by dragging fields in Power BI.

Next: Formatting, Themes and Report Design.