AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is the Unit of Experimentation Standardized?
The way that we built or or metric framework m is we built it initially around user ideas and what we called a aai unique visitor ideas. Unique visitor ideas are basically am it's like a coalescie of user and visitor. So if we can match a device to a user, then we match it to a user. If we can't, then leave it as as lica cookier of israel eh. And so pretty much every experiment in the early days was run against a unique visitor or a user idea. It just depends n if they're logged in. I think tis always a challenge for data scientists to decide, should i run mi on sessions or like customer id