Types of Graphs: 13 Charts and When to Use Each

There are many types of graphs, and each one is built to answer a different kind of question. A bar chart compares amounts. A line chart follows change over time. A histogram shows how values are spread out. Pick the wrong type and the same numbers can look confusing or even misleading. Pick the right one and the answer is obvious at a glance.

In everyday speech, people use “graph” and “chart” for the same thing, and this guide does too. It covers thirteen types: bar, grouped bar, stacked bar, pie, donut, line, histogram, box plot, dot plot, scatter plot, Venn diagram, timeline and word cloud. For each one you will see what it shows, one rule for using it well, and one case where you should choose something else.

Every example below is drawn live in your browser. Each has an Open in button. Click it and the chart opens in the matching free tool with the data already filled in, so you can change the numbers and download your own version. All numbers in the examples are sample data.

How to Choose a Graph: Start With the Question

Do not start with the chart. Start with the question you want the reader to answer. Then match it to a type.

• Which is bigger? Compare categories with a bar chart.
• What is it made of? Show parts of a whole with a pie or donut chart, or a stacked bar.
• How did it change? Follow a trend with a line graph.
• How are the values spread out? Use a histogram, a box plot or a dot plot.
• Are two things related? Plot them on a scatter plot.
• What do these groups share? Draw a Venn diagram.
• What happened, and in what order? Build a timeline.
• Which words come up most? Make a word cloud.

Keep that list open while you read. The rest of the guide goes through the types in the same order.

Graphs That Compare Categories

Bar Chart

A bar chart shows categorical data as rectangular bars. The length of each bar is proportional to the value it stands for, and the bars can run vertically or horizontally. A vertical bar chart is sometimes called a column chart, according to Wikipedia’s entry on the bar chart.

Each bar is a different category, so the chart is about comparison. One axis lists the categories and the other axis shows the measured value. When the categories have no natural order, you can arrange the bars any way you like. Sorting them from highest to lowest makes the ranking easy to read, and Wikipedia calls a bar chart sorted that way a Pareto chart.

One rule: start the bars at zero. Wikipedia notes that a non-zero baseline can make similar values look deceptively different. The exceptions are measures such as temperature in Celsius or Fahrenheit, where zero is somewhat arbitrary.

Choose something else when you want to show how values spread out inside one group. That is a job for a histogram, which comes later in this guide.

William Playfair is widely credited with inventing the bar chart. His Commercial and Political Atlas used one to show the exports and imports of Scotland for a year around 1780 to 1781. Many sources credit him, which is why you will often see him named.

Open this bar chart in the bar graph maker

Grouped Bar Chart

A grouped bar chart, also called a clustered bar chart, puts two or more bars next to each other inside each category. Each color stands for one series, such as one term or one year. Use it when you want to compare the series with each other inside every category.

One rule: keep the number of series small. Claus Wilke, in his open textbook Fundamentals of Data Visualization, points out that reading a color takes more effort than reading a position, because you have to match each bar to the legend. With two or three series that effort is small. With seven it is not.

Choose something else when the total matters more than the parts. That is where the stacked bar comes in.

Stacked Bar Chart

A stacked bar chart piles the series on top of each other in one bar, so the height of the whole stack shows the combined total. Wilke says stacking is useful when the sum of the individual amounts is itself a meaningful amount. Counts of paperbacks, hardcovers and e-books add up to the total number of books, so stacking them makes sense.

One rule: put the series you care about most at the bottom. Only the bottom segment sits on the axis. Wilke notes that stacked bars have the problem that the bars are hard to compare because they have different baselines. Our full guide to the stacked bar chart explains when it works and when it misleads.

Choose something else when you need to compare the middle or top parts across categories. Use grouped bars instead.

Graphs That Show Parts of a Whole

Pie Chart

A pie chart divides a circle into slices, and each slice is one share of the whole. Its history goes back to William Playfair’s Statistical Breviary of 1801, which is the earliest known example. Both Wikipedia and Stephen Few (2007) give that date.

Pie charts have a well-known weakness. Stephen Few argued in 2007 that a pie encodes values by the area and the angle of its slices, and that people compare areas and angles less accurately than positions and lengths. That is why a bar chart usually reads better when you want to compare categories with each other.

They do have a strength. Wikipedia reports that if the goal is to compare one slice with the total, and the share is close to 25 or 50 percent, a pie can beat a bar chart. Wilke agrees that pies work well for simple fractions such as one half, one third or one quarter.

One rule: keep the slices few. With many sections, colors repeat or look alike and the chart becomes hard to read. As a rule of thumb of our own, not a rule from a study, we suggest five or six slices at most. Group the small ones into “Other”.

Choose something else when the reader needs to rank the parts. Bars line up on a common baseline, so the ranking is easy to see. If a pie is still the right call, build one in the pie chart maker.

Donut Chart

A donut chart is a pie chart with a blank center. The center can hold information about the whole, such as a total. Its purpose is the same as a pie chart: to show proportions.

One rule: use the hole for one short number or label, and nothing else.

Choose something else when you have more than a handful of parts. The donut inherits every weakness of the pie.

Graphs That Show Change Over Time

Line Graph

A line graph shows data points ordered by their x value, usually time, joined by straight segments. It looks like a scatter plot, except that the points are ordered and connected. That is why it is the usual choice for a trend over time.

Wilke describes the idea well. In a time series the dots are spaced evenly along the x axis and have a defined order. Connecting neighboring points with lines makes that order easy to follow, and the result is called a line graph.

Playfair did not invent the line graph, but Stephen Few writes that his work popularized it as a way to show values across time.

One rule: put time on the horizontal axis, from oldest to newest, with equal steps between the points.

Choose something else when the x axis is not ordered. If the categories are just names, connecting them with a line suggests a trend that does not exist. Use bars.

Open this line graph in the line graph maker

Graphs That Show How Values Are Spread Out

Some questions are not about one number per category. They are about a whole set of numbers: how they cluster, how far they range, and whether any of them stand out. Three graphs answer that.

Histogram

The NIST/SEMATECH Handbook says a histogram summarizes the distribution of one numeric variable. It shows the center of the data, its spread, its skewness, the presence of outliers and the presence of more than one mode. The most common form splits the range into equal-sized bins, called classes, and counts the points in each bin.

In a histogram the boxes touch each other, and the vertical axis shows frequency or relative frequency. OpenStax, in Introductory Statistics 2e, gives a rule of thumb: use a histogram when the data set has 100 values or more. That is a practical guide, not a law, and the example above uses 30 values so the chart is easy to follow.

One rule: try more than one bin size. There is no single best number of bins, and different bin sizes reveal different features of the data.

Choose something else when each bar is a separate category. In a histogram each bin is a range of one numeric variable, while in a bar chart each bar is a different category. Some authors suggest gaps between the bars of a bar chart to keep the two apart. The histogram maker can build bins for you with a rule such as Sturges or Freedman-Diaconis.

Box Plot

A box plot summarizes a data set with five numbers. NIST describes the box as running from the lower quartile, the 25th percentile, to the upper quartile, the 75th percentile. A line marks the median, and the box holds the middle 50 percent of the data.

OpenStax builds a box plot from five values: the minimum, the first quartile, the median, the third quartile and the maximum. In the example above the first quartile is 2, the median is 7 and the third quartile is 9, with a minimum of 1 and a maximum of 11.5. When a chart marks outliers with dots, the whiskers no longer reach the minimum and maximum. In the common version described by NIST, a point beyond 1.5 times the interquartile range from the box is treated as a mild outlier and drawn separately.

One rule: use box plots to compare several groups side by side. One box is fine, but the real value shows up when you line up three or four.

Choose something else when you want to see the shape of the data. A box hides whether the values cluster in two peaks. A histogram shows it.

Dot Plot

Wikipedia describes a dot plot as a chart of data points on a simple scale, typically drawn as filled circles. It suits small to moderate data sets and highlights clusters, gaps and outliers while keeping the actual numbers. Unlike a histogram, which only shows how many values fall in each bin, a dot plot keeps every value: one dot is one observation.

One rule: keep it small. Wikipedia says that with roughly 20 to 30 points or more, a stemplot, box plot or histogram may be more efficient, because dots get cluttered.

Choose something else when you have a large data set. A stemplot is another option worth knowing about. OpenStax calls it a quick way to graph data that gives an exact picture of the data. Our dot plot maker is built for the small-set case.

Graphs That Show a Relationship

Scatter Plot

A scatter plot reveals a relationship, or association, between two variables. NIST puts the variable you suspect is related on the horizontal axis and the response on the vertical axis. It helps you ask whether the two are related, whether the relationship is linear, whether the spread of Y changes with X, and whether there are outliers.

When the dots follow a roughly straight path, NIST notes that a linear regression model might be appropriate. The line drawn through the dots is the best-fit line. OpenStax explains that r squared, the square of the correlation coefficient, tells you the share of the variation in y that is explained by x through that line.

One rule: a pattern in a scatter plot shows that two things move together. It does not show that one causes the other.

Choose something else when your x values are categories. That is a bar chart.

Graphs for Sets, Sequences and Words

Venn Diagram

A Venn diagram uses overlapping closed curves, usually circles, to show all possible logical relations between a finite collection of sets. Points inside a curve are elements of that set, and the overlap of two curves is their intersection. In a Venn diagram the curves overlap in every possible way. A diagram that leaves some relations out is an Euler diagram, so a Venn diagram is a special case of one.

One rule: each item goes in exactly one region. If you cannot decide where something belongs, the sets need clearer definitions.

Choose something else when you have numbers, not groups. Our Venn diagram maker draws two or three circles.

Timeline

A timeline is a list of events in chronological order, usually drawn as a bar labeled with dates. Our timeline tool spaces the events evenly in the order you enter them rather than to scale, and the dates are free text such as “Week 3” or “Spring 2025”. So it is best for a story or a plan, not for measuring gaps.

One rule: lead each entry with the event and keep the note to one sentence.

Choose something else when the exact size of the gaps is the point. Plot the values on a line graph. If you want ready-made layouts, see our timeline templates.

Word Cloud

A word cloud, also called a tag cloud, is a visual representation of text data. It shows the importance of each word with font size or color, and it is often used to summarize free-form text or keywords.

One rule: use it for a quick first look at a text, such as survey answers.

Choose something else when the reader needs the numbers. A word cloud shows that a word is big, not by how much. A bar chart of the top ten words gives exact counts.

A Quick Chooser

Types of Graphs in Statistics

In a statistics class, most graphs are about distributions and relationships. Those are the histogram, the box plot, the dot plot, the stemplot (also called a stem-and-leaf plot) and the scatter plot. They all work on numbers rather than on categories. If your class names a graph that is not in this guide, such as a frequency polygon, the ideas above still apply: decide what question it answers and then check what it hides.

The four statistics tools on this site are the histogram maker, the box plot maker, the dot plot maker and the scatter plot maker. Each one takes pasted numbers and draws the chart in your browser, with no sign-up.

Common Mistakes

1. Cutting the bars. A bar axis that starts above zero makes small differences look large. Start at zero.
2. Too many slices. A pie with a dozen slices cannot be read. Switch to bars.
3. Using a bar chart for a distribution. If the horizontal axis is a number line split into ranges, you want a histogram, where the boxes touch.
4. Connecting categories with a line. Lines imply order and change.
5. Leaving out labels. Give every chart a title, label the axes, and add units.
6. Using color as the only signal. Add labels or values, so the chart still works in print and for readers who cannot tell certain colors apart.
7. Showing a relationship as proof of a cause. A scatter plot shows that two things move together, nothing more.

Other Types You Will Meet

This guide sticks to the thirteen types that cover most school and office needs. You will also see area charts, radar charts, heat maps and Gantt charts. We do not cover them here, and the tools on this site do not draw them. When you meet one, ask the same three questions: what does it show, what does it hide, and would a simpler type do the job?

Try It With Your Own Data

The fastest way to learn a chart type is to build one. Open any example above, replace the numbers with your own, and download the result as a PNG or a clean SVG. Nothing you type leaves your browser.