Bar Chart: When to Use It and How to Read It

Best for: Comparing one number across a handful of categories, such as sales by region or users by plan.
Avoid when: The categories are points in time with many steps (use a line), or you need to show a distribution (use a histogram).
Data you need: One categorical column and one numeric column. Optional second category for grouped or stacked bars.

A bar chart compares an amount across categories by drawing each amount as a bar of matching length. That is the whole idea, and it is why the bar chart is the chart most people read correctly on the first try. Our eyes are good at comparing lengths that share a baseline, and a bar chart gives them nothing else to do.

That also makes it the chart people abuse most, because it seems too simple to get wrong. Most of the problems below are small, and every one of them is easy to fix.

When to use a bar chart

Reach for bars when you have one number per category and the question is "which is bigger, and by how much?" Revenue by product, survey answers by option, page views by country: all bar charts.

Bars still work when you add a second category, for example revenue by product for two years. Beyond two or three series, though, the chart turns into a forest and readers start hunting for colors in a legend. At that point small multiples (one small bar chart per series) usually read faster.

Bars are a poor fit for long time series. Thirty daily bars side by side push the reader to compare heights one by one, while a line shows the shape of the change at a glance. They also fail for distributions, where a histogram is the right tool.

Sort the bars

Unless the categories have a natural order (age bands, months, survey scales from "strongly disagree" to "strongly agree"), sort the bars by value. Alphabetical order is almost never what the reader needs. It answers "where is Peru?" when the real question is "who is ahead?"

Fig. 1 The same seven values. On the right, the ranking is visible without reading a single number, and one highlighted bar shows where to look.

Sorting also makes a highlight work. Once the bars are in order, coloring one of them tells the reader exactly where to look and how it ranks against the rest.

Start the axis at zero

A bar shows its value through its length. If the value axis starts at 88 instead of 0, a bar for 94 ends up twice as long as a bar for 91, even though the real gap is about 3%. Anyone who compares the bars, which is the whole point of the chart, gets the wrong answer.

Fig. 2 Three values between 91 and 94. The truncated axis on the left turns a 3% difference into what looks like a doubling.

When the small difference really is the story (a conversion rate moving from 3.1% to 3.4%, say), don't fake it with a cut axis. Plot the change itself, or use a dot plot. Dots mark a position rather than a length, so a zoomed axis doesn't distort them.

Horizontal or vertical

Use horizontal bars when the category names are long or when you have many categories. The labels sit on the left and read like a list, with no rotated text. Vertical bars (often called column charts) suit a short list, or categories with a left-to-right order such as years or age groups.

Rule of thumb: if you are tempted to rotate the axis labels by 45 degrees, flip the chart to horizontal instead.

Grouped bars and their limits

Grouped bars put two or more series side by side within each category. They work when the reader needs to compare the series inside each group, like this year against last year in each region.

Fig. 3 Two series per category is about the limit. Each pair is easy to compare, and the gaps between groups keep the regions apart.

Comparing across groups is harder. In the example above, telling whether the South's 2024 bar is taller than the East's means skipping over two other bars. If cross-group comparison matters more, put each series in its own panel, or show the difference between the two years as a single bar.

Labels, gridlines and color

A few decisions do most of the work:

Label the bars directly when there are fewer than about a dozen. A number at the end of each bar is easier than tracing a gridline back to the axis, and once every bar is labeled you can usually drop the axis.
Keep gridlines light or remove them. They are reference marks, not part of the data.
Use one color for all bars, and a second color only for the bar you want to talk about. Rainbow bars make the reader look for a meaning the colors don't have. The color guide covers this in more depth.
Write the takeaway in the title. "Peru grew faster than any other market" tells the reader what to look for; "Growth by country" leaves them to work it out.

Common mistakes

Mistake	Why it hurts	Fix
Axis that doesn't start at zero	Lengths no longer match values	Start at zero, or use a dot plot
Alphabetical order	Hides the ranking	Sort by value
3D bars	Perspective changes apparent height	Always flat
A different color per bar	Suggests categories mean something	One color plus one highlight
Too many series per group	Readers lose track of the legend	Small multiples, or fewer series
Thick bars with thin gaps	Bars blur into a block	Gap of roughly half a bar width

A good bar chart looks almost boring. When the data is clean, sorted and labeled, the comparison stands out by itself and the reader never has to decode anything.

Questions

Should a bar chart always start at zero?

Yes. A bar encodes its value through its length, so cutting the axis makes small differences look large. If the differences you care about are small compared with the values, show the change itself or switch to a dot plot, which does not rely on length.

Horizontal or vertical bars?

Horizontal bars are easier when category names are long or when there are more than about seven categories, because the labels can be read without tilting your head. Vertical bars feel natural for a short list or for ordered categories like age bands.

What is the difference between a bar chart and a histogram?

A bar chart compares separate categories, so the bars have gaps and can be reordered. A histogram shows how one numeric variable is distributed across ranges, so the bars touch and their order is fixed.