Thomas, CRO consultant

I've been running A/B tests for ecommerce clients since before Optimizely had a free tier. When someone pitches me a book about upselling, my first question is always the same: "Does it understand that statistical significance requires sample size, or does it just throw around percentages like confetti?"

This book mostly passes that test.

The author clearly knows the difference between a directional finding and a properly powered experiment. When he cites Uber Eats' 18–25% uplift, he notes that it varied by market. When he discusses ActiveCampaign's 12% lift from renaming a tier, he specifies it was in the European market during Q1 2022, on a localized pricing page. That granularity matters. Most "data-driven" business books would have just said "renaming your tiers can boost conversion by double digits," full stop. This one gives you enough detail to evaluate whether the finding applies to your context.

My main critique is that the book is better at diagnosis than prescription. It tells you what to look at — your timing, your labels, your catalogue schema — but it doesn't give you a rigorous testing framework. The practical implications at the end of each chapter are sensible but light. "Move one upsell module and compare the next fifty transactions" is fine advice for a small merchant, but it's not a substitute for a real experimentation methodology. I would have liked a chapter on testing infrastructure: how to structure a hypothesis, how long to run a test, how to handle segment interactions. That's the piece most businesses get wrong, and the book doesn't really address it.

That said, I've already used the chapter on the user's state of mind with two clients. The idea of suppressing offers during active task flow and surfacing them at natural boundaries — after a save, after an export — is something I've tested before, but the book gave me a clearer conceptual framework for explaining it to stakeholders. The section on time-of-day effects in travel booking is also useful; I'm designing a test around evening vs. morning offer framing for a hotel aggregator right now.

Recommended, with the caveat that you'll need to bring your own statistical rigor. The book provides the ideas; you provide the experimental design.