To test the consistency of a percentage metric like

conversion_percent

over time, the most relevant anomaly detection monitors in Elementary would typically be: - average: Monitors the average value of the column over time, which is useful for tracking the central tendency of a percentage metric. - variance and standard_deviation: These monitor the variability of the metric over time, helping you detect unusual fluctuations or instability in your percentage values. - min and max: These can help you catch unexpected drops or spikes in your percentage metric. You would generally not use

null_count

,

sum

, or

zero_count

for a percentage metric unless you specifically want to monitor for missing values or the occurrence of zeros. A typical configuration for monitoring consistency might look like this:

yaml models:   - name: your_model     columns:       - name: conversion_percent         tests:           - elementary.column_anomalies:               column_anomalies:                 - average                 - variance                 - standard_deviation                 - min                 - max

This setup will help you detect anomalies in the average, variability, and range of your

conversion_percent

metric over time, which aligns with monitoring its consistency. For more details on supported monitors, see the documentation on [column anomalies configuration](https://docs.elementary-data.com/data-tests/anomaly-detection-configuration/column-anomalies) and [column anomalies test](https://docs.elementary-data.com/data-tests/anomaly-detection-tests/column-anomalies).