We've found that different experimentation goals need different statistical approaches:
* Classic experiments when you need to learn precisely from costly variations
* Safe rollouts when you want to ship confidently with automatic rollbacks
* Bandits when you need to pick winners from many quick-to-build options
Article: https://blog.growthbook.io/flavors-of-experimentation-in-growthbook/
#OpenSource #Experimentation #Statistics #DataScience
