Rule out a hidden cause before calling a correlation a driver

Two numbers move together on my chart and my manager wants to know if one is causing the other.

Beginner to intermediateRead 5 minDoing: 30 to 45 minutes for a first pass, quicker once the query is saved6 steps

What you'll have at the end

A short note naming the possible third cause you checked, before the chart is called evidence that one thing drives the other.

Before you start

You need

A chart already showing two of your own numeric columns moving together, plus query access to the individual rows behind it.

Not covered

This does not run a formal significance test or build a full causal model; it only checks for one plausible hidden cause before the chart gets called proof of anything.

Leans on

6 steps

Where to go next

Compare two groups' averages with a t-test

for turning the eyeballed group gap into a real significance test rather than a comparison by eye

Checked 26 Aug 2026

Part of the Descriptive Statistics cookbook