Use a bar graph when your data falls into distinct categories, like product names or survey responses. Use a histogram when your data is numerical and continuous, like test scores, ages, or income levels. The distinction comes down to what kind of data you have and what question you’re trying to answer.
The Core Difference: Categories vs. Continuous Numbers
Bar graphs display categorical data. Each bar represents a separate, distinct group: favorite streaming services, sales by region, votes per candidate. The categories don’t flow into one another. “Netflix” doesn’t blend into “Hulu” the way the number 25 blends into 26.
Histograms display numerical, continuous data. Each bar represents a range of values (called a bin), such as home sale prices from $100,000 to $150,000, or ages from 30 to 39. The whole point of a histogram is to show you the shape of your data’s distribution: where values cluster, how spread out they are, and whether the pattern skews left or right. If you want to understand the underlying distribution of a numerical data series, a histogram is the right tool.
Why the Bars Look Different
The most visible difference is the gaps. Bar graph bars have spaces between them, signaling that each category is separate and unrelated. Histogram bars touch, with no gaps at all. This isn’t just a style choice. Because continuous data flows from one value to the next without interruption, the adjacent bars reflect that continuity visually. A gap between “20–29 years old” and “30–39 years old” would falsely suggest that no one can be 29.5.
There’s also a difference in how you can arrange the bars. In a bar graph, you’re free to reorder categories however you like: alphabetically, by size, by importance. Sorting bars from tallest to shortest often makes comparisons easier. In a histogram, the order is fixed. The x-axis represents a number line, so the bins must follow their natural numerical sequence. Moving a bin would destroy the shape of the distribution, which is the entire reason you made the chart.
Real-World Examples for Each
A quick way to decide: ask yourself whether your x-axis labels are names or numbers.
- Bar graph: Number of votes each candidate received in an election. The candidates are distinct categories.
- Bar graph: Website traffic broken down by source (search, social media, direct, referral).
- Bar graph: Percentage of people who prefer various streaming services.
- Histogram: Distribution of home sale prices across a city. You want to see where most prices cluster.
- Histogram: Test scores for a class of 200 students. You want to spot whether scores bunch around the middle or spread evenly.
- Histogram: Age distribution within a population, showing how many people fall into each decade of life.
- Histogram: Income levels across a workforce, revealing whether most employees earn near the median or whether a long tail stretches toward high salaries.
The Gray Area: Ordinal Data
Some data sits between purely categorical and truly continuous. Ordinal data has categories that follow a natural order but aren’t measured on a continuous scale. Think of survey responses like “strongly disagree, disagree, neutral, agree, strongly agree,” or age ranges that have already been grouped into buckets like “18–24” and “25–34.”
For ordinal data, a bar graph is typically the better fit, but you should keep the bars in their natural order rather than sorting by height. Rearranging “agree” before “strongly disagree” just because it’s taller would strip out the meaningful progression. Some visualization guides treat ordered bar charts and histograms as interchangeable here, but the key question remains: are you showing a distribution of raw measurements, or comparing pre-defined groups? If the bins already exist in your data (as with survey scales), use a bar graph. If you’re creating the bins yourself from continuous measurements, that’s a histogram.
What a Histogram Reveals That a Bar Graph Can’t
A histogram’s real power is showing distribution shape. When you plot continuous data as a histogram, you can immediately spot patterns that a bar graph would hide. Is the data symmetric, with a bell-curve shape? Does it skew right, with a long tail of high values (common with income data)? Is there a bimodal pattern, with two separate peaks suggesting two distinct subgroups?
None of these patterns are visible in a bar graph because categories have no inherent spatial relationship. Placing “Marketing” next to “Engineering” on a bar chart of department headcounts tells you nothing about the shape of a distribution. But placing the $200K–$250K salary bin next to the $250K–$300K bin tells you exactly how quickly the frequency drops off at higher incomes.
Choosing Bin Widths for Histograms
If you’ve decided a histogram is the right choice, your next decision is how wide to make the bins. Too few bins and you’ll oversimplify the data, hiding important features. Too many and the chart becomes noisy, making it hard to see the overall pattern.
Most software handles this automatically, but the logic behind it is straightforward. One common rule (Sturges’ rule) sets the number of bins to roughly 1 plus the base-2 logarithm of your sample size. For 1,000 data points, that gives you about 11 bins. Another approach (Scott’s rule) calculates bin width based on the spread of your data and the sample size, producing wider bins when data is more variable and narrower bins when you have more observations. A third option, proposed by Freedman and Diaconis, uses the interquartile range instead of standard deviation, making it more resistant to extreme outliers pulling the bin width too wide.
In practice, start with whatever your software suggests, then adjust. If the histogram looks blocky and featureless, try more bins. If it looks like static, try fewer. The goal is to reveal the data’s shape without introducing visual noise.
A Common Mistake With Bar Graphs
One pitfall worth knowing: bar graphs can mislead when they display averages rather than counts. Research from Wellesley College found that about one in five people misinterpret bar graphs showing averages. These readers treat the top of the bar as the maximum value in the data rather than the midpoint. They sketch nearly all individual data points below the bar’s tip, which is mathematically impossible for an average.
The takeaway is practical. A bar works best when it represents a single, concrete number, like a count (150 hospital beds) or a specific quantity ($5.75). When a bar represents an average of many underlying values, it hides the spread and variation in those values. If you need to show both the central tendency and the distribution, a histogram or a chart that displays individual data points is a better choice.
Quick Decision Guide
Ask yourself three questions. First: is my data made up of names and categories, or measurements and numbers? Categories get a bar graph; measurements get a histogram. Second: do I care about comparing groups, or understanding how values are distributed? Comparison points to a bar graph; distribution points to a histogram. Third: does the order of my x-axis matter? If you could rearrange the labels without losing meaning, it’s a bar graph. If rearranging would destroy the pattern, it’s a histogram.