Kohavi and Thomke explain how controlled online experiments (A/B tests) enable data-driven product decisions at scale. Covers experimental design, statistical significance, common pitfalls like peeking at results too early, and how companies like Microsoft and Booking.com run thousands of experiments annually.