I want to stress that we need to have a relatively concrete list of problems the changes are meant to address which will help us develop the questions we want to ask which will help us design a schema for collecting the data that answers those questions.

Here's what I've got written so far:

• changes are meant to decrease the bounce rate
• changes are meant to reduce time spent on the portal

There was a third one that Moiz mentioned that I couldn't quite catch.

Wider research document at https://meta.wikimedia.org/wiki/Research:Portal_experiments#Initial_data_collection
Basic schema; https://meta.wikimedia.org/wiki/Schema:WikipediaPortal

@MSyed your review would be appreciated!

@Ironholds this looks good! We should get someone else to review the basic schema too I think. Someone who has done this before and can help us identify anything we're missing.

@EBernhardson would be a good person to talk about the viability.

Thinking about it there are a couple more fields we should/could add:

1. Something analogous to accept_language. If it largely agrees with what wiki they end up with, that points to there being some advantage in automagically changing what links we give people based on their browser's language settings.
2. A 'cohort' field, NULL by default, that we can use for A/B testing.

@Ironholds: Those fields were added, right?

Da!