Assign user IDs not visitor IDs to experiment variation

Avatar

Avatar

matt-gardner

Avatar

matt-gardner

matt-gardner

31-10-2018

I want to be able to run AB tests that consistently drop a user into a specific (randomized) experiment experience - no matter which browser they sign in on.   The experience is post sign in - so essentially I want to randomly assign user ids to test and control - as opposed to visitor ids.   Is this possible?  I cannot find out how to do this in the documentation.  Can someone point me in the right direction - I may be looking for the wrong thing!

Thanks,

Matt

View Entire Topic

Avatar

Avatar

ryanr701

Employee

Avatar

ryanr701

Employee

ryanr701
Employee

19-03-2019

Hi matt-gardner,

The specific use case you outlined can actually be achieved a few different ways with Adobe Target. As Shruti mentioned device co-op can bridge cross device scenarios. Though it can even help before login. Mbox3rdPartyId is another way that will allow you to share profile data across devices post login. Here's the main help doc: Real-time profile syncing for mbox3rdPartyID​. However, the specific scenario you mention could also be achieved with a simple profile parameter and an audience rule (assuming there is a logic to the assignment of id's) and an XT activity (rather than AB activity). Though it can still be treated/evaluated as an AB test.

For example:

  1. Userid 10001 with a Target ID (aka tntid) of abc logs in to your site. In their Target mbox request you can pass profile.userid=10001 to Target.
  2. If you have an audience rule where all visitors with a profile.userid ending in odd digits (or ending with 0-5 or some other pattern) I can assign her to experience A in an XT activity.
  3. Now if userid 10002 comes along (tntid: def) my audience rule for the even digits will assign him to experience B.
  4. Now if userid 10001 logs in on device 2 (so a different tntid: ghi) you still are able to pass the profile.userid of 10001 to Target because you talking about a post login scenario, and since the audience rule checks against the profile.userid Target knows to place this visitor into experience A.

Caveat is while only 2 userid's have enter the test, Target will report 3 visitors: 2 in experience A (tntid: abc, ghi), and 1 in experience B (tntid: def). If you care about the reported visitor counts then device co-op will be a better option. IF you simply care about maintaining the visitors experience across devices you simply need to pass the id you are keying off of and have a logical way of splitting the values into the different experiences/groups.

Since you've assigned exp A to odd ids and exp B to even ids they will split pretty evenly as long as the userid values are actually random. So while I've set this up as an XT activity in Target it is still effectively being an AB test.

Hope that helps you understand some of the other options.