What this book covers
Ron Kohavi, Diane Tang and Ya Xu address the practical and statistical work behind trustworthy online experiments. For agencies, this is relevant when a result must withstand scrutiny beyond a dashboard's winner label. Use it to improve test design, measurement and interpretation, while considering whether the account has enough data for the question.
Bring it to your work
Draft a test brief with assignment method, primary metric, guardrails, expected traffic, decision rule and reasons a result could be invalid. PMC editorial prompt, not a quoted book exercise or tested outcome.
Access and review scope
Paid book; price varies by retailer and format. Full book not read. Review based on front matter, contents and preface. Website A/B methods need adaptation for auction or geographic incrementality tests.
About this reading guide
This guide draws on the source description, contents or sample identified in the edition notes. No external review is cited for this listing.
Prepared with AI assistance from the sources identified on this page. The applications and comparisons are Paid Media Collective suggestions, not claims of full-book reading. An external review does not verify every claim in a book.
Another book to consider
Experimentation Works →
Choose this when the obstacle is organisational support for testing rather than test mechanics.
Edition & source notes
The reference below identifies the material checked for this listing. It may differ from the format offered by the main book link.
- Reference edition
- First published 2020; paperback
- Format
- paperback
- Reference edition date
- 2020
- ISBN-13
- 9781108724265
Publisher or author description reviewed. Sources checked . Book not read in full.
Reading context: Evergreen discovery.
Selected public evidence; full book not read.
Access: Paid book; price varies by retailer and format.
Paid book; price varies by retailer and format.
Full book not read. Review based on front matter, contents and preface.
Website A/B methods need adaptation for auction or geographic incrementality tests.
Source facts and reading prompts prepared with AI assistance. The named authors and linked publishers retain credit for the original work. Source verification is separate from an assessment of current platform tactics.
Take the idea off the page.
What would make us distrust a positive result before changing the campaign?
Talk with the collective
