Power BI’s built-in anomaly detection is primarily designed for numerical time series data, meaning it detects anomalies based on fluctuations in numerical values over time (e.g., sales, revenue, counts). It does not natively support anomaly detection on purely categorical fields such as product types or customer segments without aggregating them into numeric metrics.
However, you can still analyze anomalies related to categorical data by aggregating them into meaningful numeric measures—such as count of transactions per category, average revenue per segment, or sales volume by product type—and then applying anomaly detection on those numeric trends. This approach helps you uncover unusual patterns within or across categories.
Best practices for implementing this include:
-
Creating measures that quantify categorical behavior (e.g., "Units Sold by Product Category")
-
Using a line chart with a time axis to visualize trends over time
-
Applying the "Find anomalies" feature on these aggregated values
-
Using slicers or filters to drill down by category for more focused analysis
For advanced anomaly detection on non-numeric or mixed-type data, you may consider exporting data to tools like Python, R, or Azure Machine Learning, and then integrating results back into Power BI via custom visuals or Power Query.
