Thomas uses the launch of Instagram’s feed ranking as a working example to talk through issues in quantifying network effects, while exploring unusual A/B testing techniques such as country-level tests, testing on balanced graph partitions, and author-side experiments.