Statistical Thinking for the 21st Century
Source: https://statsthinking21.org/ ↗
Poldrack is a Stanford neuroscientist who wrote an open-source statistics textbook because he was tired of what was available for his own students.
The book covers the fundamentals — probability, hypothesis testing, regression — with modern emphasis on effect size, Bayesian reasoning and reproducibility.
For product direction it is the most rigorous free option on the shelf: it does not shy away from formalism, but it does not hide behind it either.
The book is continuously updated, which is itself a small lesson about how serious educational material travels in the open-source era.
Pair it with Wheelan for the intuition layer and Gill for a four-page Bayesian warm-up.
Central argument
Poldrack argues that statistical practice in science has been systematically distorted by an over-reliance on null hypothesis significance testing and p-values, and that a more honest, reproducible approach requires centering effect sizes, uncertainty quantification, and Bayesian reasoning. The book does not treat these as advanced electives but as foundational to how any quantitative claim should be built and evaluated. The implicit thesis is that statistical illiteracy is not a deficit of calculation skill but of conceptual framing — and that framing can be taught from first principles without sacrificing rigor.
Critique
The book is written for graduate students in behavioral and neural sciences, and that origin shows: the worked examples lean heavily on experimental designs typical of academic research rather than on observational, high-velocity, or instrumented data environments that define product contexts. A CPO or data practitioner managing A/B test debt, metric gaming, or causal inference in logged behavioral data will occasionally have to do the translation work themselves. The open and continuously updated format, while admirable, also means there is no stable edition to cite or assign — a minor but real friction for teams trying to build shared vocabulary around a fixed reference.
Why it matters for product
Product teams routinely make consequential decisions on statistically underpowered experiments, conflating statistical significance with product significance — exactly the confusion Poldrack's emphasis on effect size and practical significance is designed to dissolve. A CPO who internalizes the Bayesian framing will also be better equipped to challenge the binary pass/fail culture around A/B testing, pushing instead toward posterior reasoning about magnitude and direction of effect. The chapter-level treatment of reproducibility maps directly onto the organizational problem of experiment registries, holdout discipline, and preventing HARKing in roadmap post-mortems.