Hi, I want to do A/B test on my page, but don't want to store any session identifiers in cookies or ...
b
Hi, I want to do A/B test on my page, but don't want to store any session identifiers in cookies or local storage. I made a script that hashes the user IP and user-agent and returns a number from 0-7 (group_id), which I send with events to my data store. The issue is that now GrowthBook only sees 7 users (it probably ignores 0, but I can shift the ids by 1), but what is worse is that it has assigned 6 of those values to Control and only 1 to Variation 1 (I have configured 50% split). What would be the best way to fix this, or is Growth Book unsuitable for this case? I don't want to track individual users or their sessions because of GDPR.
I basically need to force set variation_id based on my group_id and ignore the hashing that GrowthBook does.
h
We don't really support custom hashing / forced variations in the way you're imagining. Our notion of forced variation is for testing/QA purposes and bypasses the normal experiment flow, skips tracking callbacks, etc. Can you use the IP directly as your hash attribute instead of trying to force a bucket? Also should note that bucketing on IP can lead to bias because its very common for many users to have the same IP address.
Alternately you could skip our SDK entirely and do your own experimentation implementation using your own hashing etc, and calling your event tracker directly. Then use GB to analyze the experiment data only instead of fully implementing in GB.
b
Thanks, skipping the SDK is a great idea, I have just a single experiment with a single flag, so it will be only a few lines of code.
I know about the bias, I can track number of visits in each group. By the way, do you know about some resources how to make A/B testing GDPR compliant? What does and doesn't require consent from users?
Most of what Google finds are ads for some products