How to Accelerate Experimentation

The flywheel paper is an attempt to help companies get started and sort of accelerate experimentation. The first experiment that you run should not be something that's aimed at having a lot of impact, but one where half the company thinks it won't work. You want whatl we call a counter intuitive result which teaches people to predict what an experiment is going to do. I think optimization is an important part of any scientific programme or experimentation program. And i would argue that the role of the central team is not to run experiments, but to look at that fly wheel and to figure out where is currently the most friction,. What which of these steps is reduced it is blowing us down?