Hey all - I wanted to share some material that I created for a talk I gave last month at Columbia for a seminar on Online Experimentation: The Fundamentals of Building and Maintaining an Experimentation Platform. It's in shitty-first-draft mode but I came across GrowthBook while doing prep, and want to potentially expand it in a way that allows a reader to touch and poke it and understand how experimentation capabilities (and associated potential problems) grow. I'd be very happy to get y'alls feedback.

Hey Paul, just took a look through. I like the emphasis on metrics as statistics free and the numberlines idea is kind of where GrowthBook is now, but we want to make it more obvious/manipulable by experimenters that use our platform.

Polish for sure is one thing I want to add. But beyond the static documentation, the main next leap here is to have "real" experiments running inline with the documentation to put something tangible around the various themes and topics. Imagine - at the most basic scenario - clicking a button and rapidly seeing an experiment execute. Behind the scenes it's a straightforward simulation with some treatment effect. That generates results and that's the first baby step in understanding. From there, successive tweaks and changes are provided to demonstrate the various ways in which things can get better or worse. Main point is to minimize/eliminate the ramp up cost of "running" the experiment to learn what it's really all about.