Outliers – Too Low Values

Purpose

This measure is used to identify unusually low values within a dataset. Detecting such outliers helps highlight potential pricing errors, data quality issues, or exceptional business situations that may require further investigation.

Business Context

Outliers can significantly affect performance analysis and decision-making. This measure uses Tukey’s test, a robust and widely accepted statistical method, to determine whether values fall significantly below the expected range. The calculation is typically performed at an aggregated level and then used as a reference threshold to identify unusually low observations.

Definition

:question_mark:

The measure calculates a threshold below which values are considered abnormally low. Using Tukey’s test, values are classified as too low if they fall below:

Q1 − 1.5 × IQR

Where:

  • Q1 = First Quartile (25th percentile)

  • IQR = Interquartile Range (Q3 − Q1)

This can also be expressed as:

2.5 × Q1 − 1.5 × Q3

Business Formula

Threshold for Too Low Values = 2.5 × Q1 − 1.5 × Q3

Agent Expression

2.5 * PERCENTILE_CONT(Metric, 0.25) - 1.5 * PERCENTILE_CONT(Metric, 0.75)