What is correlation vs causation in email performance?

Still have a question, spotted an error, or have a better explanation or a source we should cite?

Correlation means two things move together. Causation means one thing directly caused the other. The distinction sounds academic but it comes up constantly in email analysis, and getting it wrong leads to optimizing for the wrong things.

Common correlations that aren't causation

Open rate and revenue. Campaigns with higher open rates often produce more revenue. But higher open rates don't cause the revenue. Both are usually caused by a third factor: a more engaged audience segment or a more relevant offer. Sending to a cold, unengaged list won't produce revenue even if you somehow forced open rates up.

Subject line length and performance. Short subject lines sometimes correlate with higher open rates. But the relationship isn't consistent across all senders and audiences. Shorter lines don't cause opens. The correlation, when it exists, is partly because shorter lines tend to be clearer and more direct. The clarity causes the improvement, not the character count.

Send day and open rate. Tuesday gets cited as the best day for email. The data that produced that figure came from averages across many senders and industries. Your audience may behave completely differently. Their engagement pattern and your specific correlation may be stronger on Thursday or Sunday.

How to distinguish them

The reliable way to establish causation in email is a controlled A/B test where you change exactly one variable. If open rate goes up after changing the subject line, and everything else (audience, send time, content) stayed the same, you have reasonable evidence that the subject line caused the change. Without that control, you have a correlation that might or might not be meaningful.

The practical rule: treat correlations as hypotheses, not conclusions. Test them before making permanent program changes based on them.

Contributors

Who worked on this answer

Every name links to their profile. Every company links to their site. Real people, real accountability.

Ask an AI · tailored to your setup

Help Me Test Whether This Pattern Is Real

I just read the Email Almanac entry on correlation vs causation in email performance. Help me figure out whether the patterns I'm seeing in my email data are real causes or just coincidences worth testing. Walk me through: 1. Whether the performance pattern I've observed could have a third-variable explanation 2. How to design a controlled test that would actually establish causation for what I'm seeing 3. Whether I have enough data for a valid test or whether I'm pattern-matching on noise 4. What variables to control for in my next A/B test --- My details (fill in what applies): - Pattern I've observed: [e.g., "emails sent on Tuesday perform better" / "shorter subjects get more opens" / other] - List size: rough number - Whether I've A/B tested this pattern: yes / no - ESP or sending platform: Mailchimp / Klaviyo / Brevo / other - How many campaigns this pattern is based on: number - Whether the pattern holds across segments or only some: all / only some / unsure

Edit the yellow boxes, then send to the AI of your choice.