Part I: Foundations

Published

December 27, 2025

Modified

October 14, 2026

Keywords

A/B Testing, Causal Inference, Causal Time Series Analysis, Data Science, Difference-in-Differences, Directed Acyclic Graphs, Econometrics, Impact Evaluation, Instrumental Variables, Heterogeneous Treatment Effects, Potential Outcomes, Power Analysis, Sample Size Calculation, Python and R Programming, Randomized Experiments, Regression Discontinuity, Treatment Effects

Get news about the book. Leave your email here.
Found an error or have a suggestion? Submit feedback here.
Follow the book’s page for new content and follow the author.

You might think causality is another fancy tool or a twist on statistics, but it isn’t. It’s a mindset, a way to structure how we think about change and what drives that change. This first part focuses on how you approach data, with less emphasis on formulas.

Yes, I’ll introduce the necessary statistical concepts and methods, but with the goal of giving you a new perspective. You’ll start seeing data differently. Here’s a roadmap of what we’ll cover and why it’s important:

Chapter 1: Data, statistical models, and ‘what is causality’

Key message: Not all data science is about prediction. Machine learning helps forecast what might happen. Causal inference asks, “What would’ve happened if we had done something differently?”

  • Learn the key terms: Population, sample, outcome, covariate (a variable that provides context), parameter, and estimate. Think of them in familiar settings like product rollouts, A/B tests, and business metrics.
  • Distinguish three essential concepts: The estimand is the effect you want to measure, the estimator is the method you use, and the estimate is the number you obtain.
  • Linear regression isn’t magic: When used appropriately and under the right assumptions, it can be a useful starting point for causal analysis.
  • Correlation is useful, but causation informs decisions: Correlation tells you how variables move together. Causation tells you what would happen if you intervened and changed something.

The payoff: Move beyond guessing. Whether it’s a feature launch or a promo, you’ll learn to reason about what would’ve happened otherwise.

Chapter 2: Biases, causal frameworks, and causal estimands

Key message: Data alone doesn’t cut it. It shows patterns, but not what caused them. That’s where you (and causal thinking) come in.

  • Omitted variable bias: You think your feature caused more purchases, but maybe those users were already heavy buyers. Confounding is sneaky.
  • Selection bias: Sometimes the way users get into the treatment group is already skewed - and that messes with your conclusions.
  • Potential outcomes framework: A way to think about “what would’ve happened if…” Since you can’t see both realities for the same person, you need to use smart comparisons.
  • ATE (impact for everyone), ATT (impact on those treated), and LATE (impact on the persuadable) - each one answers a different question with a business use case.

The payoff: Catch mistakes before they catch you. You’ll learn to spot flawed conclusions and set up studies that people can trust.

Chapter 3: Planning and designing credible causal analyses

Key message: Causal analysis doesn’t start with a method. It starts with a sharp question. And a plan to avoid misleading results.

  • Learn how to rewrite vague stakeholder asks - like “did our new feature help?” - into clear hypotheses with defined treatments, outcomes, and comparisons.
  • Changing your hypothesis after seeing the results is tempting… and dangerous. You’ll learn how to stay credible.
  • Choose the right setup. Randomized tests if you can, smart observational strategies if you can’t (DiD, IVs, RDDs… I’ve got you).
  • Don’t run analyses in a vacuum. Ask: what decision does this support? What action does it unlock?
  • You’ll see a full example: how to run the analysis, check for fair comparisons (balance), interpret the output, and explain the uplift to your product manager.

The payoff: Credibility by design. You’ll learn to transform vague stakeholder requests into rigorous, auditable studies that withstand scrutiny and produce trusted results.

Why this pays off

Mastering these foundations puts you ahead of most data professionals. You’ll gain the vocabulary to spot questionable conclusions in executive meetings and the confidence to communicate what the data can actually say.

I use realistic examples so you can see how this causal mindset drives business growth. These basics are prerequisites for everything that follows.