Correlation

Measure pairwise correlations between variables using Pearson, Spearman, and Kendall methods.

Overview

The Correlation tool calculates how strongly variables in your dataset are related to each other. It computes three correlation coefficients — Pearson, Spearman, and Kendall — between your chosen targets and variables, helping you identify linear and non-linear relationships.

The dataset must have at least 5 epochs to run correlation analysis.

How to Run

Open Correlation Metrics from the navigation.
Select Modeling Combinations — one or more Country/Region/Grouping triplets.
Select Targets — the variables to correlate against (defaults to the Sales group).
Select Variables — the candidate variables (defaults to the Price group).
Click Get Correlation Metrics.

Reading the Results

Table View

A sortable, filterable table with columns: Target, Variable, Combination, Pearson, Spearman, and Kendall. Expand a row to see additional detail.

Matrix View

A heatmap showing pairwise correlations between all selected variables at a glance. Stronger correlations appear as more intense colors.

Correlation Methods

Method	Measures	Best for
Pearson	Linear relationship	Continuous variables with a linear trend
Spearman	Monotonic relationship (rank-based)	Ordinal data or non-linear but monotonic trends
Kendall	Concordance (rank-based)	Small samples or data with many tied values

Tips

High Pearson but low Spearman may indicate outliers are driving a spurious linear correlation.
Use correlation results alongside Feature Importance to distinguish between correlated variables and truly predictive ones.
Check for multicollinearity (high correlations between predictor variables) before model building.