Solved problems. Proven solutions.

Small sites rarely have enough traffic for perfect statistical isolation, so use controlled change logs, comparable periods, and multiple signals instead of declaring victory from a few clicks. 1. Define the hypothesis and primary metric before making the change. 2. Change one meaningful group of variables at a time when practical and document the exact dep…