What is data visualization? Pick the chart by the question

Data visualization turns numbers into charts so the pattern jumps out. The trick is to pick the chart by the question you are answering, not by how it looks.

Short answer: data visualization means showing numbers as a picture, a chart, so a pattern is seen at once instead of read row by row. Choose the chart by your question: a line for change over time, bars to compare, a donut for parts of a whole, a histogram for how values spread.

A table of 200 numbers holds the answer, but nobody sees it. The same numbers as a line show in one second that sales dropped in March. That is the whole job of a chart: make the one thing that matters visible.

Nothing loads from YouTube until you press play. Then Google receives your IP address and may set its own storage. Narrated with an AI-generated voice. Privacy

Which chart answers which question?

Your questionChartWhy
How did it change over time?LineThe slope shows the direction.
Which is bigger?BarLengths side by side are easy to compare.
What share is each part?Donut or pie, with two to five partsSlices of one whole.
How are the values spread?HistogramCounts per range show the shape.
Do two numbers move together?ScatterEach dot is one item, on two scales.
Are we on target?Bullet chartThe bar against a target mark.

Why start with the question?

Most bad charts are picked by habit: a pie because it looks friendly, a line because the last report had one. A pie with twelve slices or a line through categories with no order hides the answer. Name the question first ("which region sold most?"), and the chart follows from it.

What else decides the chart?

The shape of your data. How many categories you have, how many points in time, how many numbers per item. Five regions fit a bar chart; fifty fit a sorted bar chart or a table. Two years of months fit a line; two numbers fit a sentence.

What makes a chart easy to read?

  • One message per chart. If it says two things, make two charts.
  • Sorted bars, unless the order means something (months, sizes).
  • Colour for meaning only: one highlight, the rest grey. Palettes everyone can read covers colour blindness.
  • Direct labels instead of a legend when there is room.

More on single charts: the histogram, the pie chart and bar or column.

Where can you check this?

Can a tool pick the chart with you?

Chart Compass does this with your own numbers. Paste a table, say what the chart should show, answer two or three questions, and it draws the charts that fit, ranks them with the reason for each, and names them in Power BI, Tableau and Excel. Its chart map holds 50 charts grouped by 11 things a chart can say. Which chart should I use? goes through the questions in more depth.

The tool

Do it in your browser.

Data

Chart Compass

Paste your data, say what the chart should show, answer two or three questions, and see the charts that fit drawn with your own numbers, with their names in Power BI, Tableau and Excel.

Keep reading

More from the blog.

Every post