How to Find Outliers in Tableau

Finding an outlier in your data can feel like discovering a secret - it might be a mistake that's quietly throwing off all your averages, or it could be a crucial insight into your most valuable customer or a wildly successful campaign. This article will show you three practical ways to find and analyze outliers in Tableau, from simple visual checks to more powerful statistical calculations.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

What Are Outliers (And Why Should You Care)?

An outlier is a data point that is noticeably different from the other observations in a dataset. Imagine you run an e-commerce store where most products sell for between $20 and $100, but one accidental transaction was recorded for $10,000. That $10,000 sale is an outlier.

But not all outliers are mistakes. They can fall into two main categories:

• Errors or Noise: These are often the result of data entry mistakes (like the $10,000 typo), measurement errors, or system glitches. They can seriously skew metrics like your average order value, making your performance seem much better or worse than it actually is.
• Genuine Extreme Values: These are legitimate but rare events. It could be an A-list celebrity buying your entire product line, a corporate client placing an unusually large bulk order, or a single blog post going viral and bringing in a month's worth of traffic in one day. These aren't errors, they are valuable signals.

Pinpointing these outliers is important. By identifying them, you can clean up your data for more accurate reporting and, more importantly, learn from the exceptions to your normal business patterns. What drove that massive sales spike? Who is that hyper-valuable customer? Finding outliers is the first step toward answering these questions.

Method 1: Spot Outliers Visually with Charts

The fastest and most intuitive way to start looking for outliers is to visualize your data. A sudden spike in a line chart or a lonely dot far from the crowd in a scatter plot can tell you almost instantly where to direct your attention. Tableau is built for this kind of quick visual discovery.

Using a Scatter Plot

A scatter plot is fantastic for finding outliers when you're comparing two different measures, like sales versus profit or ad spend versus conversion rate. Outliers will appear as points that are far away from the main cluster of data points, breaking the general trend.

Here’s how to set one up:

1. Connect to your dataset in Tableau.
2. From the Data pane, drag a measure like Sales to the Columns shelf.
3. Drag another measure like Profit to the Rows shelf. Tableau will generate a single point representing the total sales and profit.
4. To see individual data points, drag a dimension that represents a unique level of detail, such as Customer Name or Order ID, onto the Detail card in the Marks pane.

Your view will now be a cloud of points, with each point representing a customer or order. Any points floating far away from the central group are your outliers. You can hover over them to instantly see the details and begin your investigation.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

Using a Box-and-Whisker Plot

The box plot (or box-and-whisker plot) is one of the most effective tools for identifying outliers in a single measure across different categories. This chart type was practically designed for spotting values that fall outside the typical range.

Here’s what a box plot shows:

• The Box: Represents the middle 50% of your data. The bottom of the box is the first quartile (Q1) and the top is the third quartile (Q3).
• The Median Line: The line inside the box marks the median of your data.
• The "Whiskers": These lines extend from the box to the highest and lowest values that are still within a standard range (specifically, within 1.5 times the Interquartile Range, or IQR).

Tableau automatically defines any data point that lands outside the whiskers as an outlier. Creating one is simple:

1. Drag a dimension, like Product Category, to the Columns shelf.
2. Drag a measure, like Sales, to the Rows shelf.
3. In the top right corner, click on the "Show Me" panel and select the box-and-whisker plot.

Tableau will instantly render the plot, displaying any outliers as individual dots above or below the whiskers for each category. This makes it incredibly easy to see which product categories, for example, had unusually high or low sales events.

Method 2: Use Standard Deviation to Flag Outliers

Visual inspection is great, but sometimes you need a more systematic, mathematical way to define what counts as an outlier. One of the most common statistical methods is using standard deviation. The general idea is to flag any data point that falls unusually far from the average of all the data points.

A common rule of thumb is to consider any point more than two or three standard deviations from the average as an outlier. Here's how to build this logic directly in Tableau using calculated fields.

Step 1: Create the Standard Deviation Calculation

We'll use a calculated field known as a "Z-Score," which measures how many standard deviations a specific point is from the average. To do this, you’ll first need to build a view, like SUM(Sales) by Customer Name.

Once you have your view ready, create a new calculated field:

• Open the "Analysis" menu and select "Create Calculated Field."
• Name the field Z-Score.
• Enter the following formula. This formula uses window calculations, which compute across the data in your current view:

(SUM([Sales]) - WINDOW_AVG(SUM([Sales]))) / WINDOW_STDEV(SUM([Sales]))

This formula subtracts the average of all sales in the view (WINDOW_AVG) from each individual sales total (SUM([Sales])) and then divides it by the standard deviation of all sales (WINDOW_STDEV). Drag this new 'Z-Score' field onto your view to see the score calculated for each customer.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

Step 2: Create a Field to Flag the Outliers

Now that you have a Z-Score for each customer, you can create a simple boolean (True/False) calculation to flag the ones that exceed your threshold.

• Create another calculated field and name it Is Outlier? (Std Dev).
• Enter this formula:

ABS([Z-Score]) > 2

The ABS() function ensures we catch outliers on both the high and low ends (e.g., a Z-score of -2.5 or 2.5). The > 2 flag marks anything more than two standard deviations away. You could raise this to 3 to apply an even stricter definition of what is an outlier.

Now, drag your Is Outlier? (Std Dev) field to the Color card in the Marks pane. Your outliers will instantly be highlighted, giving you an automatic, data-driven way to separate them from the rest.

Method 3: A More Robust Approach with IQR

The standard deviation method is great, but it has one weakness: it can be influenced by the very outliers you're trying to find! If you have a few extremely large values, they'll pull up the average and stretch out the standard deviation, potentially masking other outliers.

The Interquartile Range (IQR) method is a more robust alternative because it relies on the median, which isn't affected by extreme values. The logic is the same one used by the box plot: any value below Q1 - 1.5 * IQR or above Q3 + 1.5 * IQR is an outlier.

Here’s how to build this in calculated fields.

Step 1: Calculate Q1, Q3, and the IQR

Again, we will use table calculations that compute based on the data in our current view. We need to create a few fields. With your view of Sales by customer still active, create three new Calculated Fields:

• Q1:

WINDOW_PERCENTILE(SUM([Sales]), 0.25)

• Q3:

WINDOW_PERCENTILE(SUM([Sales]), 0.75)

• IQR:

[Q3] - [Q1]

Step 2: Define the Upper and Lower Boundaries

With the core components built, we can now define the limits outside of which our data points are considered outliers.

• Lower Bound:

[Q1] - 1.5 * [IQR]

• Upper Bound:

[Q3] + 1.5 * [IQR]

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

Step 3: Flag the Outliers

Similar to the last method, the final step is to create a true/false calculated field to identify the outliers in the dataset.

• Is Outlier? (IQR):

SUM([Sales]) < [Lower Bound] OR SUM([Sales]) > [Upper Bound]

When you drag this Is Outlier? (IQR) field to the color card, you may discover that it flags more or fewer customers as outliers compared to the standard deviation method. This method allows you to confidently and systematically find the outliers in your data so your team can focus on discovering why they are different from the rest.

You've Found Outliers… Now What?

Finding the outliers is just the first step. The more important work is deciding what to do next. Throwing them away isn't always the right move - their story is often where the most significant insights live.

Investigate the Context

Before doing anything else, figure out why a data point is an outlier. Use Tableau's "View Data" feature to look at the underlying record. Is it a typo? Does it belong to a particular campaign, region, or product line? Are a bunch of outliers coming from the same source? Digging in helps you distinguish between an error and an opportunity.

Decide Whether to Keep, Filter, or Segment

Your investigation will guide your next action:

• Filter/Remove: If it's clearly an error (e.g., a customer's age is listed as 200), it’s safe to right-click and filter it out of your main analysis to keep a clear view of typical company performance.
• Keep: If it's a legitimate, massive sale, you absolutely want to keep it. This represents a real event that you need to learn from. However, when reporting on averages and totals, it can easily distort and provide a misleading view of standard operations.
• Segment: This is the most powerful option. Create two groups from your data: one for "Outliers" and one for "Standard" customers. Create new views to compare their behavior side by side. What's different about their demographics, their purchase history, or how they found your brand? This is where real data-driven strategy begins.

Final Thoughts

In this walkthrough, we've gone over various methods to help you identify outliers including visual discovery using scatter and box plots, or through systematic calculation of standard deviation and interquartile range. This gives your entire team the power to quickly recognize patterns and spot potential issues without deep statistical modeling knowledge, so they can focus on why these outliers happened and how to use these insights to drive growth and stay ahead.

At Graphed, we help with just that. Using our tools, you can gain deeper insights by simply asking questions in plain English - no calculated fields and steep learning curves. We connect all your sales and marketing data in one place and let you build real-time dashboards and reports by asking natural language prompts like "Show traffic from US, Canada, and UK." Gain immediate insights from a single unified view without stitching together spreadsheets and CSV exports.