Skip to main content

Anomaly detection (preview)

Anomaly detection helps you enhance your line charts by automatically detecting anomalies in your time series data. It also provides explanations for the anomalies to help with root cause analysis. With just a couple of clicks, you can easily find insights without having to slice and dice the data.

Since this feature is in preview, you will need to first turn on the feature switch by going to File > Options and Settings > Options > Preview feature and make sure Anomaly detection is turned on:

You can enable Anomaly detection by selecting the chart and adding the “Find Anomalies” option in the analytics pane.

For example, let’s look at this chart showing Revenue over time. Adding anomaly detection automatically enriches the chart with anomalies and the expected range of values. When a value goes outside this expected boundary, it is marked as an anomaly.

 

This experience is highly customizable. You can format the anomaly’s shape, size, and color, and also the color, style, and transparency of the expected range. You can also configure the parameter of the algorithm. If you increase the sensitivity, the algorithm is more sensitive to changes in your data. In that case, even a slight deviation is marked as an anomaly. If you decrease the sensitivity, the algorithm is more selective on what it considers an anomaly.

Besides detecting anomalies, you can also automatically explain the anomalies. When you select the anomaly, Power BI runs an analysis across fields in your data model to figure out possible explanations. It gives you a natural language explanation of the anomaly, and factors associated with that anomaly, sorted by its explanatory strength. In the example below, you can see that on August 30, Revenue was $5187, which is above the expected range of $2447 to $3423. You can open the cards in this pane to see more details of the explanation.

You can also control the fields that are used for analysis. For example, by dragging Seller and City into the Explain by field well, Power BI  restricts the analysis to just those fields. In this case, the anomaly on August 31 seems to be associated with a particular seller and particular cities. The visual in the card shows the spike in the revenue for this seller on August 31. You can use the “Add to report” option to add this visual to the page.

The report consumers can view anomalies and explanations after the creator publishes the report to the service. Viewing anomalies and explanations in the report consumption experience in Power BI Mobile (iOS, Android, and Windows) will be supported soon.

Please try out this visual while it is in preview. We greatly appreciate any feedback in terms of what you liked about the feature and how we can improve it!  If you have any feedback for the team, please comment on our community post.