Overview

The Sifflet Distribution Change monitor is a field-profiling monitor that detects changes in the volume percentage of categories within a dataset. It helps identify distribution anomalies, either through dynamic detection or static comparison.

How To

How Does It Work

The Sifflet Distribution Change monitor calculates the volume percentage of each distinct value within the monitored column and compares the resulting distribution either against a defined threshold or historical reference data.

The column can contain values of any type, as long as the cardinality remains within the supported limit.

📘

Limit on the number of categories

Sifflet limits the number of categories to 1200. Above this threshold, the monitor run returns a Needs Attention status.

How to Configure

Find the Distribution Change template in the Field Profiling category.

Two Options

After selecting the columns to monitor, you can enable the following options:

• Fail if new category appears — The monitor fails if a previously unseen category (or combination of categories) is detected.
• Fail if category disappears — The monitor fails if an expected category (or combination of categories) is missing from the data.

Monitoring Type: Dynamic, Static, Relative

You can choose between dynamic, static, and relative thresholds.

Dynamic Mode

The monitor fails if the statistical test finds any anomaly based on the previous trends. You do not need to define a threshold, but you need to set a sensitivity.

Static Mode

The monitor uses a predefined threshold as a reference. It fails if any category’s volume percentage increases or decreases beyond the specified threshold compared to the reference distribution.

Relative Mode

In Relative mode, the monitor evaluates changes in category volume percentages using one of two comparison methods:

• % change — Compares the percentage increase or decrease relative to the previous value.
  Example: A change from 10% to 20% is considered a 100% increase.
• % points change — Compares the absolute difference in percentage points.
  Example: A change from 10% to 20% is considered a 10 percentage point increase.