Unexpected Changes

Last modified: March 23, 2022

This article provides an overview of how Auto Insights detects anomalies, outliers, big movements, and changes.

Auto Insights helps you to uncover insights that you might otherwise have missed, either because it would be impossible to analyze all factors manually, or simply because you didn’t know where to look! 

As you upload data to Auto Insights, it scans your data to find unexpected changes. If found, Auto Insights surfaces insights that may require further investigation in cards. These insights are automatically surfaced in Discover and Missions. 

What types of insights does Auto Insights look for?

Auto Insights looks for Anomalies and Outliers.


Anomalies are values that sit outside the expected range. Auto Insights detects anomalies by calculating the expected result (based on historic trends) in comparison to the actual result. This comparison factors in previous trends and considers seasonality. To detect anomalies, we use algorithms such as STL, S-ARIMA, ARIMA, Random forest and PCA.

Auto Insights only searches for anomalies on a monthly basis and will only search for anomalies if you upload more than seven months of data.


Example: Sales for Cosmetics tend to decrease between June and July each year. Between June and July 2019, sales for Cosmetics increased by 6.52%. Auto Insights considers this behaviour to be an anomaly, as the actual result (an increase in sales) differs from the expected result (a decline in sales).

Read more here on the different types of anomalies that Auto Insights can surface out.


Outliers are values that have experienced large growth or decline when compared to their peers for the same period. For example, comparing the growth of different departments within an organisation, or even individual performance in a team. This helps you to benchmark performance so you can spot elements that have experienced unusually high or low growth.

Auto Insights detects outliers by calculating the average range of growth across a segment in your data (e.g. the average growth of sales for all departments in an organisation). Auto Insights considers an outlier to be a value (e.g. a department) that sits outside this average range of growth. We use a number of algorithms to detect outliers, for example, the interquartile range (IQR).

In contrast to anomalies, Auto Insights can detect outliers on as little as two periods of data (2 days, 2 weeks or 2 months). In combination, anomalies and outliers gives you a full view of unexpected changes across different time periods.


Example: In this example, the growth range for all departments was -11% to 50%. Seafood department is considered an outlier because its growth of -33% puts it outside the overall average range. 


Related Articles
Getting Insights From Auto Insights

Types of Anomalies

FAQ on Unexpected Changes

Was This Page Helpful?

Running into problems or issues with your Alteryx product? Visit the Alteryx Community or contact support. Can't submit this form? Email us.