More combinations do not automatically make a better test. In this interview, I describe a mistake that is easy to make early on: splitting the account into lots of audiences, devices and locations before finding a useful creative direction. The result can be a lot of spend spread across comparisons that are difficult to learn from. Here is a way to put that into practice: 1. List what you are trying to learn. Distinguish a creative question from an audience, placement or geography question instead of mixing them all into one giant experiment. 2. Choose the most useful first comparison. Ask which decision would make the biggest difference to the next round of work, given the resources you actually have. 3. Keep the structure understandable. You should be able to explain why each test exists and what would change if it performs differently from the baseline. 4. Check whether the budget supports the plan. A large number of variations can leave very little evidence for any single comparison; do not assume more rows mean more learning. 5. Expand after a clear lesson. Carry a useful creative direction into another test, document the change and keep the unanswered questions visible. Use this as a planning checklist before launching another batch. The aim is enough focus to interpret the result, not a magic number of ad sets. - @mr.paidsocial