Stop Measuring Correlations. Start Proving Causation.

Why your marketing analytics is probably lying to you — and how to fix it.

If you found this useful, you can support my independent research here: https://www.paypal.biz/vector1research

Your email campaign shows recipients spent 40% more than non-recipients.

The data is clear. The p-value is tiny. Leadership wants to 10x the email budget.

But here’s what the data doesn’t tell you: customers who open emails are already more engaged. They would have spent more anyway.

The email might have had zero effect. And you’d never know from the correlation alone.

This is the correlation trap. It has caused more wasted marketing budget than any other analytical error.

The $2 Million Mistake

I watched it happen.

A marketing team scaled a “winning” campaign based on strong correlation data. Millions in spend. Months of execution.

The result? No incremental lift. The “effect” they saw was entirely driven by customer selection bias.

They were measuring correlation. They thought they were measuring causation.

The Methods Exist

Here’s the thing: we know how to solve this problem.

Economists, epidemiologists, and tech companies have developed rigorous methods to separate correlation from causation:

• A/B testing — when you can randomize

• Difference-in-differences — when you have pre/post data

• Regression discontinuity — when there’s a sharp cutoff

• Propensity scores — when you’re stuck with observational data

These aren’t new. They’re just not widely used in marketing.

The Guide

I spent the last several months writing the guide I wish I’d had.

The Practitioner’s Guide to Causal Marketing Analytics

It covers:

Part I: Foundations
Why dashboard metrics mislead. The correlation trap (with code showing exactly how it happens). Thinking causally about marketing.

Part II: Statistical Modeling
Regression that works for MMM. When ML predictions don’t equal causal effects. Forecasting vs. causal inference.

Part III: Causal Inference Methods
A/B testing done right (sample sizes, peeking, sequential testing). Quasi-experiments (DiD, RD, IV, synthetic control). Propensity score methods (matching, IPW, doubly robust). Causal discovery algorithms.

Part IV: Simulation
PyCausalSim framework for building intuition. Validating your methods before trusting results.

Every chapter includes production-ready Python code. 57 modules total, all tested.

Who This Is For

Marketing analysts tired of reporting correlations they can’t defend. Data scientists moving into marketing. Marketing leaders who need to make real budget decisions.

This isn’t for beginners. You need basic Python and statistics knowledge. But you don’t need a PhD — everything is explained from first principles.

Get the Guide

Three tiers depending on your needs:

• Essential ($97) — The complete guide

• Professional ($297) — Guide + all Python code

• Team ($997) — Everything + consultation + team license

Link: https://currywave832.gumroad.com/l/alhesq

— -

Brian Curry is the founder of Vector1 Research, specializing in causal inference for marketing applications.