Guide When to cut a campaign/ad/geo/zone/offer (using statistics and scary math stuff)...

[Jimmy]

I've rewritten and reposted this guide, so check out the new version here:

When to Cut a Zone That Has Conversions (Using Statistical Analysis)

afflift.com

Spoiler: If you still want to read the original post, it's here (with a few pictures missing)

If math was never your thing then this post might be a bit difficult to digest at first, but hopefully I’ll be able to keep it simple enough that you won’t drown straight away and can come back after a couple of Red Bulls and try and force these concepts down. I think they’re important to understand in this game, at least to some degree and I spent quite a bit of time early on banging away at this until it made sense and it's really helped a lot.

It’s going to be unavoidable that this post is going to discuss some mathematical concepts that have long and complicated (scary) names – things like “statistical significance” and “binomial confidence intervals”. But don’t worry, that’s about as tricky as it’s going to get. I’ll leave talking about “stochastic discrimination” and “random decision forests” to the guys from @Kintura which seems to have some pretty advanced AI capabilities that members like @DEADZ have been talking about:

A Tracking Experiment (& Journey) To Find The "Best" Tracker - DEADZ

afflift.com

Reading