Error Bars: A Guide to Visualizing Uncertainty

Not all data points are equal. Some are exact, while others are estimates or averages. When we present our data, it’s essential to be upfront and clear about any uncertainty. When we don’t, we leave room for people to misinterpret our charts or make poor decisions based on our data. But by using error bars, we have a straightforward tool to show how reliable our data is. 

Error bars give everyone a quick way to see how precise and trustworthy our data is. In this guide, we’ll cover what error bars are, how to interpret them, the different types you’ll see, and practical steps for adding them to your own charts. Let’s make your data more transparent and more informative to help your people make better decisions.

What are error bars?

Error bars are basically a graphical way to show how much your data might vary. They’re used on charts to indicate the uncertainty or possible error in a reported measurement. Think of them as showing a range of plausible values for a data point, not just a single, discrete value. They typically appear as little T-shaped lines extending from the central point on a graph.

What do these lines actually tell you? They can show several different things, depending on what you want to say about your data:

• Standard deviation: This shows how spread out the individual data points are from the average.
• Standard error: This tells how precise the average of your sample is likely to be compared to the true population average.
• Confidence intervals: This gives a range of values where you can be reasonably certain the true population mean lies.
• A specific range: These are the minimum and maximum values observed in a data set.

Keep in mind that error bars are there to supplement your chart, not replace the main data points. They add a layer of context, helping viewers understand how much confidence they can place in the data presented. By showing uncertainty, you give a more honest and complete picture.

When (and why) to use error bars

Adding error bars provides context and transparency to your charts, changing how people interpret the data. But not every chart needs them. Use error bars when your data points are estimates rather than exact figures.

Errors bars are a great way to add clarity to:

• Experimental results: In scientific research, measurements are often repeated to ensure accuracy. Error bars can show how much these measurements might vary, indicating the precision of the experiment.
• Forecasts and projections: When you predict future values, like sales or stock prices, there’s always a degree of uncertainty. Error bars can show the potential range of outcomes, from optimistic to pessimistic scenarios.
• Sample-based data: Surveys and polls gather data from a sample of a larger population. Error bars present the margin of error, helping the audience understand how well the sample results might reflect the whole population.

Analysts use error bars to promote transparency, increase confidence in precise findings, and enable more accurate comparisons. By showing uncertainty, error bars reveal the limitations of your data and provide a way to assess whether differences are statistically significant or just due to chance.

Sometimes, error bars aren’t appropriate. Avoid them with categorical data that lacks variability or when data is highly noisy, as very large error bars can clutter your chart and obscure trends.

How error bars work: the mechanics

To use error bars effectively, you should understand how they’re structured and calculated. Error bars have three components: a central value, an upper bound, and a lower bound. The method you choose to calculate these limites defines what the error bars tell you.

At a minimum, an error bar is defined by three parts:

• Central value: This is the dot or the top of the bar on your chart. It represents the mean (average), a specific measurement, or a calculated point-estimate.
• Upper bound: The top of the error bar, representing the central value plus the calculated error amount.
• Lower bound: The bottom of the error bar, representing the central value minus the calculated error amount.

The distance between the central value and the limits (bounds) is set by how you choose to calculate them, which determines what the error bars represent about your data's variability.

Common calculation methods

• Standard deviation (SD): This is one of the most common measures. It quantifies the amount of variation or dispersion in a set of data values. A low standard deviation means the data points tend to be very close to the mean, while a high standard deviation indicates that the data points are spread out over a wider range of values. Use standard deviation error bars when you want to show the spread of your raw data.
• Standard error of the mean (SE or SEM): This metric estimates how far your sample mean is likely to be from the true population mean. It’s always smaller than the standard deviation. Use standard error bars when you want to show how precise your estimate of the mean is. It’s particularly useful when comparing the means of different groups.
• Confidence intervals (CI): A confidence interval gives a range of values that you can be confident contains the true population mean, to a certain level of probability (commonly 95 percent). A 95 percent confidence interval means that if you were to repeat your measurement 100 times, you would expect the true mean to fall within that interval 95 times. Use confidence interval error bars to infer how well your sample data represents the whole population.
• Custom ranges: Sometimes, you may want to display a specific range, such as the minimum and maximum values in your data set, or a predetermined percentage of error. This approach offers flexibility but requires clear labeling so the audience knows exactly what the error bars represent.

Symmetric vs asymmetric error bars

Symmetric error bars 

These have upper and lower bounds that are equal distance from the central value. This is the standard for most error bars, such as those showing standard deviation, standard error, and many confidence intervals.

Asymmetric error bars 

These are used when the potential for error differs above and below the central value. For example, data that can’t be negative may have a lower bound that is less than the upper bound. Custom calculations or specific statistical models can also result in asymmetric intervals.

Vertical vs horizontal error bars

The orientation of error bars depends on chart type. Use vertical error bars for column and line charts to show uncertainty in y-values, and horizontal error bars for horizontal bar charts to show uncertainty in x-values. On scatter plots, you can display both types to represent uncertainty in both dimensions.

Types of error bars and their uses

Choose the error bar type that matches your analytical goal. Each kind highlights a specific aspect of your data’s variability or uncertainty. Here are the most common types and when to use them:

• Standard deviation error bars:
  
  • What they show: The spread of the individual data points in your sample.
  • When to use them: When you want to illustrate the variability within your data sample. If you measured the height of 50 people, standard deviation error bars would show how much the heights of those individuals vary from the average height. It answers the question, “How spread out is my data?”
• Standard error error bars:
  
  • What they show: The precision of the sample mean. It indicates how much the mean might vary if you were to take another sample from the same population.
  • When to use them: When you are comparing the means of two or more groups and want to highlight whether the difference between them is likely real or just due to random sampling. It helps answer, “How reliable is my estimate of the mean?”
• Confidence interval error bars:
  
  • What they show: A range where the true mean of the entire population is likely to fall. A 95 percent confidence interval is the standard, meaning you are 95 percent confident the true population mean lies within that range.
  • When to use them: When you want to make an inference about the larger population from which your sample was drawn. This is common in survey research and scientific studies. It answers the question, “What is the likely range of the true mean for the whole population?”
• Range-based error bars:
  
  • What they show: The full range of your data, from the minimum value to the maximum value.
  • When to use them: In situations where the absolute extremes are important. For example, you might use them to show the highest and lowest daily temperatures recorded over a month. Be cautious, as outliers can make these ranges very large and potentially misleading.
• Custom or user-defined error bars:
  
  • What they show: Whatever you define them to show! This could be a margin of error based on a specific business rule, a percentage of the point value, or a fixed value you want to apply across all points.
  • When to use them: When standard statistical measures don’t apply or when you want to visualize a specific, predefined uncertainty. The key here is to always explain what your custom error bars represent.

Choose the right error bar type based on your goal: use standard deviation to describe sample variability and standard error or confidence intervals to make inferences about a larger population. Clear purpose ensures accurate communication.

Design best practices and common pitfalls

Clear, well-designed error bars improve how readable your chart is and help your audience interpret the data accurately. Follow these simple best practices to improve clarity and avoid common pitfalls.

Design best practices

• Keep them visible but not overpowering: Error bars should be distinct enough to be seen but not so thick or dark that they dominate the data points. A thinner line weight in a neutral color (like gray) often works well.
• Use consistent styling: Apply the same line width, color, and cap style (the small horizontal lines at the end) to all error bars on a single chart. Consistency is key for easy comparison.
• Label or annotate clearly: Never assume your audience knows what your error bars mean. Add a note in the chart's subtitle, a caption, or an annotation explaining what they represent (e.g., “Error bars show 95 percent confidence intervals”). This is the single most important practice.
• Ensure readability: Make sure the end caps of the error bars are distinct and don’t blur into the data markers. Choosing a “cap” style over a “no cap” style generally improves readability.

Common pitfalls to avoid

• Cluttering the chart: Avoid adding error bars to every single series in a complex chart with many overlapping lines or bars. This can create a visual mess. Be selective and only show them for the most important data series or create separate charts.
• Mixing calculation types: Don’t use different types of error bars (e.g., standard deviation for one series and confidence intervals for another) within the same chart without extremely clear labeling. This can lead to serious misinterpretation.
• Hiding important data: Ensure the error bars don’t obscure the central data points. The point estimate itself is still the primary piece of information. The error bar provides context around it.
• Creating “chart junk”: Every element on your chart should serve a purpose. If the uncertainty is negligible or not relevant to your message, adding error bars might just be adding unnecessary visual noise.

By following these guidelines, you ensure error bars clarify your data rather than complicate it. Use them to guide your audience to a better understanding of your charts.

Examples and interpretation tips

Examples are the best way to see error bars in use. Here are common scenarios and tips for interpreting them.

Example: Bar chart with error bars

Consider a bar chart comparing average customer satisfaction scores for three products, each with error bars showing the 95 percent confidence interval.

How to read it:

• Product A’s bar: The top of the bar might be at a score of 85. The error bar extends from 82 to 88. This means that while the average score from your sample was 85, you can be 95 percent confident that the true average satisfaction score for all customers lies somewhere between 82 and 88.
• Comparing products: Let’s say Product B has an average score of 78, with an error bar from 75 to 81. Since the error bars for Product A (82–88) and Product B (75–81) don’t overlap, you can be confident that the difference in their satisfaction scores is statistically significant. Customers are genuinely more satisfied with Product A.
• Overlapping bars: Now, suppose Product C has an average score of 83, with an error bar from 80 to 86. Its error bar overlaps with Product A’s (82–88). This overlap suggests that the difference between their scores might not be statistically significant. It’s possible that the true average scores for both products are very similar, and the difference you see in the sample is just due to random chance.

What overlapping error bars signal

When error bars overlap across data points, it suggests that the difference may not be statistically significant. Non-overlapping error bars indicate a more likely significant difference.

Be cautious with this guideline though, especially for standard error bars. For a difference to be statistically significant at the 95 percent confidence level, the gap between standard error bars often needs to match the size of one error “arm” (from mean to bar end). With 95 percent confidence intervals, any overlap generally means no significant difference.

Error bars offer a fast visual check for statistical significance, helping you avoid drawing strong conclusions from minor or random data differences.

How to create error bars

Adding error bars to your charts is simple in most spreadsheet and data visualization tools. The overall steps are the same: prepare your data, create the chart, and add error bars. Below are steps for Excel, Google Sheets, and BI platforms.

General steps

1. Prepare your data: Your spreadsheet should be set up correctly. At a minimum, you’ll need a column for your categories (e.g., Product A, Product B) and a column for the central values (e.g., average scores). You’ll also need columns for the error values. This could be a single column if the error is symmetric (e.g., ±5) or two columns for the upper and lower bounds if it’s asymmetric. You might need to calculate these values first (e.g., using STDEV.S for standard deviation or CONFIDENCE.NORM for confidence intervals in Excel).
2. Create your chart: First, create your basic bar, column, or line chart using your category and central value data.
3. Add error bars:
   
   • In Excel: Select the data series on your chart. Go to the “Chart Design” tab, click “Add Chart Element,” then “Error Bars,” and choose an option. You can select a standard type, like Standard Deviation or Percentage. For custom values, choose “More Error Bar Options.” In the panel that opens, under “Error Amount,” select “Custom” and click “Specify Value.” Here, you can reference the cells in your spreadsheet that contain your positive (upper) and negative (lower) error values.
   • In Google Sheets: Double click on your chart to open the Chart Editor. Go to the “Customize” tab and expand the “Series” section. Check the box for “Error bars.” You can then select a type (Constant, Percent, or Standard Deviation) and enter the value. Google Sheets currently has more limited native support for custom range-based error bars compared to Excel.
4. Format the error bars: Once added, you can customize their appearance. Click on the error bars to select them, then adjust their color, thickness, and cap style to ensure they are clear but not distracting.

Using a business intelligence (BI) tool

Modern BI platforms like Domo streamline this process by automatically calculating and applying error bars to your visualizations. Features like checkboxes or drag-and-drop options make it easy to add and adjust error bars. This saves time, reduces manual errors, and lets you focus on interpreting your results.

Limitations and when to use an alternative

Error bars are useful, but they have limitations. They simplify variability to a single line, which can obscure details about a data set’s distribution.

• They can oversimplify: An error bar shows a range but doesn’t tell you anything about the shape of the data’s distribution within that range. Are the data points clustered near the mean, or are they spread out evenly? You can’t tell from a simple error bar.
• They can be hard to interpret with large data sets: On a line chart with hundreds of data points, adding an error bar to each one would create an unreadable mess of overlapping lines.
• They can be misleading: If the underlying data isn’t normally distributed (the classic “bell curve”), standard error and confidence interval calculations can be misleading.

When error bars aren’t enough, try these alternatives:

• Box plots (or box-and-whisker plots): These are excellent for showing more detail about the distribution. A box plot displays the median, the interquartile range (the middle 50 percent of the data), and the minimum and maximum values (whiskers), as well as any outliers.
• Violin plots: These are like a box plot combined with a density plot. They show the same information as a box plot but also visualize the full distribution of the data. The “violin” shape is wider where more data points are concentrated and narrower where they are sparse.
• Confidence bands: For line charts showing trends over time, a confidence band (or confidence ribbon) is a great alternative to individual error bars. It creates a shaded area around the line, representing the uncertainty at every point along the trend, which is much cleaner visually.
• Small multiples: If a single chart is too cluttered, you can break it down into a series of smaller, similar charts (small multiples). This allows you to show error bars or other details for different categories without cramming everything into one visualization.

Choose the visualization that best matches your data’s complexity and the message you want to convey.

Charting a clearer path forward

Error bars do far more than just decorate your chart. They show uncertainty and reliability, giving your audience the tools to assess how precise your data is. By using error bars, you present not just discrete numbers, but the range where those numbers may truly fall.

In this guide, we explained what error bars are, when to use them, and how to design them clearly. Always match the error bar type to your analytic goal and label them clearly so viewers know what they represent. Visualizing uncertainty builds trust and help people make better decisions.

The good news: Business intelligence platforms make it easy to quickly add error bars and other uncertainty visualizations, turning raw data into clear, transparent charts that anyone can understand.