Comparing amounts across categories
A bar chart answers one question quickly: how do the categories compare? Each bar is a category, its length a value, and the eye is superb at ranking lengths that share a baseline. Use it when the groups are names, not numbers: departments, products, countries, budget categories. Order the bars by size unless the categories have a natural order (months, age bands) that carries information of its own.
A $2,300 monthly plan split as needs 1,000, wants 600, savings 400, debt payoff 300. Sorted bars put needs first and make the structure obvious: essentials take about 43% of the total, and savings plus debt together (700) outweigh wants (600). That comparison, visible in two seconds, is the entire case for the chart type.
The zero-baseline rule
Bar lengths are judged as ratios, so the axis must start at zero. Start it at 90 and a value of 100 towers ten times higher than a value of 91 on the page, while the real ratio is 1.1. Truncated axes on line graphs are a presentational choice; on bar charts they are simply misleading, because the bar's body itself becomes the lie.
Chartjunk and category sprawl. Twenty bars with pattern fills and 3D shadows rank worse than eight plain ones. If categories multiply, group the tail into an honest 'Other' bar, and keep the labels short enough to read without tilting your head.
Bar chart or histogram?
Names on the axis: bar chart, bars apart. Number ranges on the axis: histogram, bars touching. If you can shuffle the categories without changing the meaning (departments yes, age bands no), it is a bar chart; if shuffling destroys information, treat it as ordered data and consider a histogram or line.
Frequently asked questions
Should bars be sorted?
By value, descending, unless the categories carry their own order (months, age groups, sizes). Sorting by value makes comparisons instant; alphabetical order makes readers do the work.
Why must the axis start at zero?
Because readers judge bar length as proportion. A bar drawn from 90 makes 100 look ten times 91. The geometry itself misstates the data.
Can I chart percentages instead of counts?
Yes, and often should: groups of different sizes compare fairly only as percentages. Say which you used on the chart itself, since the same bars tell different stories under each.
How many categories is too many?
Past about twelve, ranking by eye breaks down. Aggregate the small tail into 'Other' (and disclose it), or switch to a table ranked with a few bars for the leaders.