Hello team, I have a few questions regarding how t...
# announcements
Hello team, I have a few questions regarding how the GrowthBook actually works, so maybe one of you can clear up my doubts. First of all, how the information about which user was assigned to the experiment is actually stored if SDK doesn't send any user data? I saw I have a new entry in localStorage but there is no info about experiment. If the data is kept only in local storage then what happens when we clear it or the user uses another device? In that case, is it possible that he will be reassigned, but this time to a different test group, which will cause that he will see another variant of the application? My second question is about increasing the scope of the experiment. What in the situation when I initially assumed that I want to test 30% of the traffic and later I would like to increase it to, for example, 60%. Do the users who were in group A in the first phase stay in it, or the groups will be re-randomized?
👀 1
The SDKs have a tracking callback function that we call whenever someone is put into an experiment. It's up to you to use the function to store that event in your data warehouse. For example, if you're using Segment.io, you might fire an
call. Everyone does event tracking a little differently, so we decided not to be opinionated and leave that part up to you. For your question about increasing the percent of traffic, users will not be reassigned. We use deterministic hashing to assign a value from 0 to 1 to every user. If you start at 30% traffic in a 50/50 test for example, we'll define the variation ranges as
[0, 0.15]
[0.5, 0.65]
. If you later increase it to 60%, the ranges will become
[0, 0.3]
[0.5, 0.8]
. So as you can see, the ranges grow, but do not overlap or switch, so existing experiment users will continue seeing the same variation as you ramp up traffic
Thank You for your comprehensive answer. Now I understand why it's so important to connect to data warehouse.