Choosing the Right Chart for Your Data

Begin with the question

A chart is an answer to a question, so the best way to choose one is to say the question out loud. Are you comparing categories? Looking at change over time? Showing how parts make up a whole? Checking whether two measures move together? Each of these has a chart style that usually fits, and getting the match right matters more than any amount of decoration.

Before charting anything, make sure the data is in good shape. Duplicates, inconsistent spellings and stray text can quietly distort a chart, which is why data cleaning comes first.

Bar charts: comparing categories

Bars are the workhorse of everyday charting. Each category gets a bar, and the length of the bar shows the value. Because people compare lengths easily, bar charts are good for questions such as “Which product sold the most?” or “How do the regions compare?”

Some practical points:

• Start the value axis at zero. Bars encode value by length, so cutting off the bottom exaggerates differences. A bar twice as tall should mean twice the value.
• Sort the bars by size when the categories have no natural order. A ranked list is much easier to read than a random one.
• Use horizontal bars when category names are long, since the labels are easier to read.
• Keep it to a reasonable number of bars. If you have dozens of categories, consider showing the top few and grouping the rest.

Line charts: change over time

A line connects points in order, which makes it the usual choice for trends across days, months or years. Questions like “Are weekly orders rising?” or “When did the dip start?” suit a line.

• Put time on the horizontal axis, running left to right.
• Use evenly spaced time periods. If months are missing, the line can suggest a smooth trend that does not exist.
• Limit the number of lines. Two or three can be compared easily; ten become a tangle.
• A line does not need to start at zero, but be careful about zooming in so far that small wobbles look dramatic. Think about whether the vertical range is fair to the size of the change.

For a single point in time, or for categories without any natural order, a line is the wrong tool, because the connecting segments imply a progression that is not there.

Pie charts: parts of a whole

A pie shows how a total divides into shares. They work when you have a small number of slices, ideally about two to five, and the message is “this part is big, this part is small.” Hypothetically, if a household’s 100 percent of spending is split into four categories, a pie can show roughly which category dominates.

Pies have limits:

• People find it hard to compare similar-sized slices. If the differences matter, a bar chart is clearer.
• The slices must add up to a meaningful whole, such as 100 percent of something. Do not use a pie for values that overlap or are not parts of one total.
• Many thin slices are unreadable. Group the small ones into an “other” slice, and label slices directly rather than relying on a color key.

Many careful chart designers avoid pies altogether in favor of bars, which is a reasonable preference. If you do use one, keep it simple.

Scatter plots: relationships between two measures

A scatter plot puts one measure on the horizontal axis and another on the vertical axis, with a dot for each record. It is the right choice for a question like “Do days with more visitors also have more sales?” Patterns you might notice include an upward drift, a downward drift, clusters or a few isolated points.

Keep in mind:

• A visible relationship does not prove that one thing causes the other. Both could be driven by something else, or the pattern could be coincidence, especially with few points.
• Outliers stand out clearly in a scatter plot, which makes it useful for spotting data entry mistakes as well as real unusual cases. A single outlier can also pull a mean well away from a typical value.
• Label both axes with units.

Other options worth knowing

• A plain table is often best when readers need exact numbers or when there are only a few values. Not everything needs to be a chart.
• A single large number with a comparison works well on a dashboard for one key figure.
• Stacked bars can show parts of a whole across several categories, but only the bottom segment has a common baseline, so the others are hard to compare.
• Histograms show how values are spread across ranges, helpful when you want to see the shape of the data rather than category totals.

Keep any chart honest and readable

1. Give it a title that states what is shown, including the period and unit.
2. Label axes and avoid cryptic abbreviations.
3. Use few colors, and use them consistently. Do not rely on color alone for meaning.
4. Remove decoration such as three-dimensional effects, shadows and heavy gridlines, which add noise without information.
5. Show the source or note how the numbers were calculated when it matters.
6. Check the chart against the underlying numbers. A range that misses the last few rows will produce a chart that looks fine and is wrong.

A quick selection guide

These are starting points, not strict rules. If a different layout makes your message clearer, try it, and show the draft to someone unfamiliar with the data. If they understand it quickly, the chart is doing its job. For broader ideas about presenting information, the topic of data visualization covers more territory, and a simple dashboard is a good place to apply what you pick here.