← BACK TO BLOG
Causal ScienceApr 10, 20266 min read

Why Correlation Kills Growth

Most growth teams optimise for signals that look like lift. The problem is they usually aren't. Here's why correlation-driven growth compounds your mistakes — and what to do instead.

Every growth team has a dashboard full of green arrows. Conversion up. CAC down. ROAS climbing. And yet somehow, revenue growth isn't keeping pace. The metrics look healthy. The business doesn't feel it.

This is the correlation trap. And it's quietly killing more growth programs than bad creative ever will.

The signal that isn't

When you optimise a paid channel and revenue goes up the next week, the natural instinct is to credit the channel. But were those customers actually incremental? Would they have converted anyway through organic search, direct, or a referral? The correlation is real. The causation often isn't.

Last-click attribution makes this worse. It systematically over-credits the final touchpoint and under-credits the upstream work that actually created intent. You end up pouring budget into channels that harvest demand — not generate it.

What causal measurement actually looks like

The gold standard is an incrementality test: a controlled experiment where a holdout group is withheld from a treatment (an ad, a campaign, a channel) and you measure the difference in outcomes between the exposed and unexposed groups. This is the only method that isolates lift from the counterfactual.

Done well, incrementality testing regularly reveals that 20–40% of attributed revenue was non-incremental. Spend that looked efficient was actually just taking credit for conversions that would have happened anyway.

The compounding problem

Here's the part most teams miss: correlation-based optimisation doesn't just waste budget in the short term. It actively degrades your ability to grow. You scale the channels that look good on paper, starve the ones doing the real work, and over time, you hollow out the engine that was actually driving demand.

Causal measurement isn't a reporting exercise. It's a structural advantage. Every experiment that correctly identifies causal lift gives you a better input for the next decision. The system gets smarter. The returns compound.

The teams that win long-term aren't the ones with the best creative or the biggest budgets. They're the ones who build the most accurate map of what actually works — and act on it ruthlessly.

( NEXT )