DoorDash needs to run experiments to understand the impact that changes to their Assignment algorithms or product may have on their delivery drivers and consumers. Therefore, to enable fast iteration on their product and algorithms, DoorDash had to scale their experimentation platform to support more than 1,000 experiments each year. In this post, Sifeng Lin and Yixin Tang describe the cultural and technical challenges associated with increasing experiment capacity and how Doordash overcame these. They discuss creating a culture where employees fail fast and learn fast, enabled by a weekly experiment cadence and an experimentation platform that supports standardized metrics and shipping criteria and accurate measurement of interaction effects. They also review how they mitigate network effects through carefully designed switchback experiments and variance reduction techniques.