The right graph depends on what you’re trying to show, not what your data looks like in a spreadsheet. Every chart answers a specific type of question: Are you comparing categories? Showing how something changed over time? Displaying how parts make up a whole? Start by identifying your message, then match it to the chart type designed for that purpose.
Start With Your Message, Not Your Data
The most reliable way to pick a graph is to ask yourself one question: “What do I want the reader to see?” Your answer will fall into one of a few core purposes, and each purpose maps to a small set of chart types. A practical decision process looks like this:
- Identify your key message. Are you showing a trend, a ranking, a breakdown, a spread, or a connection between two things?
- Choose the chart type that matches that message. Each purpose has two or three strong options.
- Check that the chart actually communicates what you intended. Show it to someone else. If they can’t state the takeaway in one sentence, the chart isn’t working.
The sections below walk through each purpose and tell you exactly which chart to reach for.
Comparing Categories
When you want to show how different items stack up against each other, bar charts and column charts are your workhorses. The choice between horizontal and vertical comes down to something surprisingly simple: label length.
Column charts (vertical bars) work best when you have a small number of categories with short labels. Think five or six product names, age groups, or regions. The vertical layout feels natural for showing magnitude, and short labels fit neatly along the bottom axis.
Bar charts (horizontal bars) are better when your category names are long or descriptive. “Quarterly revenue by department” with labels like “Customer Success and Retention” will look cramped under vertical columns. Flip to horizontal bars, and every label reads cleanly without rotating or shrinking text. Horizontal bars also work well when you’re ranking a long list of items, since readers can scan top to bottom.
Showing Change Over Time
Line charts are the default for time-based data, and for good reason. They make trends, dips, and spikes immediately visible across weeks, months, or years. If you’re tracking a single metric over time, or comparing two or three metrics on the same timeline, a line chart is almost always the right call.
Area charts look like line charts with the space below the line filled in with color. That filled area becomes useful in two specific situations. First, when your data sets have a summation relationship, meaning you want to show how multiple parts add up to a total over time. A stacked area chart displays revenue and expenses so that the viewer can visually estimate profit by seeing the gap between the two colored regions. Second, area charts work for part-to-whole relationships that change over time, similar to a stacked column chart but with a continuous flow.
If you’re just tracking one or two lines and don’t need to emphasize volume or totals, stick with the line chart. It’s cleaner and easier to read.
Displaying Parts of a Whole
When your data adds up to 100% and you want to show how each piece contributes, pie charts are the obvious choice, but only up to a point. The rule of thumb: pie charts work well with six or fewer slices. Beyond six, the slices become too similar in size to distinguish, and the chart stops being useful.
If you have more than six categories, you have a few options. You can combine the smallest categories into an “Other” slice to get back under six. You can split the data across multiple pie charts. Or you can switch to a different chart type entirely. Treemaps handle larger numbers of categories well because they use nested rectangles of different sizes, making it easier to spot proportional differences. Stacked bar or stacked column charts also work for dense composition data, especially when you want to compare the breakdown across multiple groups or time periods.
Donut charts function the same way as pie charts (they just have a hole in the middle). They don’t add analytical value, but some people find them easier to label.
Understanding How Data Is Spread Out
Distribution questions ask: Where do most values fall? How spread out are they? Are there outliers? Two chart types dominate here, and they reveal different things.
Histograms group your data into ranges (called bins) and show how many values fall into each range. They’re great for seeing the overall shape of your data: whether it’s clustered in the middle, skewed to one side, or spread evenly. If you want a quick visual answer to “what’s typical in this data set,” a histogram is your first move.
Box plots compress the same information into a more compact form. They show the median (the middle value), the range where the middle 50% of values fall, and individual outliers as dots beyond the main box. Box plots are especially useful when you want to compare distributions across groups. For instance, comparing the spread of employee tenure across five departments is much easier with five side-by-side box plots than five overlapping histograms. Box plots also make outliers impossible to miss. In a dataset of employee tenure, a box plot might flag three people at 15 years as potential outliers, something a histogram could obscure if the bin size is too wide.
Revealing Relationships Between Variables
When you want to see whether two things are connected (does spending more on ads lead to more sales?), scatter plots are the standard tool. Each dot represents one observation plotted along two axes: one variable on the horizontal axis, another on the vertical. Patterns emerge quickly. Dots trending upward suggest a positive relationship. A random cloud of dots suggests no relationship.
Bubble charts extend the scatter plot by adding a third variable. Each dot becomes a bubble, and the bubble’s size represents a third data dimension. For example, you could plot advertising spend on one axis, sales on the other, and make each bubble’s size proportional to the number of employees in that region. This lets you evaluate three variables at once without needing a separate chart. If you only have two numerical variables to explore, a standard scatter plot is cleaner.
Matching Charts to Your Data Type
One detail that trips people up is the difference between categorical and continuous data, because certain charts only work with one type. Categorical data has a fixed set of groups: departments, countries, product types, yes/no responses. Continuous data is measured on a scale and can take any value within a range: temperature, income, time, weight.
Bar charts, column charts, and pie charts are built for categorical data. Histograms, line charts, and scatter plots are built for continuous data. Box plots sit in a useful middle ground: they display the distribution of a continuous variable, and when you place them side by side, they let you compare that distribution across categories. If you’re exploring how a continuous variable (like salary) differs across a categorical variable (like job title), side-by-side box plots are one of the most efficient options.
Mistakes That Distort Your Message
Choosing the right chart type is only half the job. A well-chosen chart can still mislead if you make a few common errors.
Truncating the y-axis is the most frequent offender. If your bar chart’s y-axis starts at 16,340 instead of zero, small differences between bars look enormous. A change from 16,340 to 16,500 murders per year is real, but a truncated axis makes it appear as though violence doubled. Always check whether your axis starts at zero for bar and column charts. (Line charts have more flexibility here, since they emphasize rate of change rather than absolute magnitude.)
Cherry-picking the x-axis is the time-based version of the same trick. By showing only the date range that supports your argument, you can make a temporary dip look like a long-term decline. Whenever possible, show enough time for the reader to judge context.
Using 3D effects on any chart type adds visual noise without adding information. The angled perspective makes it harder to judge bar heights and slice sizes accurately. Flat, two-dimensional charts are easier to read in every case.
Making Charts Readable for Everyone
Color is one of the most powerful tools in data visualization, but roughly 8% of men and 0.5% of women have some form of color vision deficiency. If your chart relies on color alone to distinguish categories, a significant portion of your audience will struggle with it.
Use labels, patterns, or different shapes alongside color so that no information is lost if someone can’t perceive the color difference. For text in your charts, aim for a contrast ratio of at least 4.5 to 1 against the background. For non-text elements like bars, lines, and map regions, the minimum recommended contrast ratio is 3 to 1 against adjacent colors. These thresholds come from web accessibility guidelines and apply whether you’re building a dashboard, a slide deck, or a report.
If you’re ever unsure whether your chart passes these tests, free contrast-checking tools are available online. Paste in your foreground and background colors, and they’ll tell you instantly whether you meet the threshold.