Hello, I have a few questions regarding the interpretation of Auto-Target in Adobe.
Suppose I want to test banner 1 (CMO) vs banner 2 vs banner 3 in a 50/50 auto-target experience.
What is my control version now? Is it banner 1 or a random allocation of banner 1, 2 & 3?
What is my Recpie B: An auto-targeted personalzied allocation of banners1,2 & 3?
Once the test runs, say I see the Recipe B has a lift.
So can I conclude that the auto-targeted version is better than control?
What if say, banner 2 was performing negativey for all visitors, but the autotargeted combination of 1,2 & 3 is still a better experience.
Is there any way to know that banner 2 underperformed and that removing it from the autotargeted combo can get even higher lifts.
Also, any suggestions or examples of when to use Auto target will also be helpful.\\
Regards,
Vivin
Solved! Go to Solution.
Topics help categorize Community content and increase your ability to discover relevant content.
Hi @vivin1 Adding comments based on my understanding for above scenario.
What is my control version now? Is it banner 1 or a random allocation of banner 1, 2 & 3?
In Auto-Target, there’s no true “static” control like a traditional A/B test. If you decide a control as banner 1, then Auto-Target may treat it as the "baseline." However, more commonly, Adobe Auto-Target will compare results based on personalization performance instead of a single fixed banner. This means it might consider a randomized version (random allocation of banners 1, 2, and 3) as the baseline.
What is my Recpie B: An auto-targeted personalized allocation of banners1,2 & 3?
Recipe B would dynamically assign either banner 1, 2, or 3 to individual users based on how the algorithm model predicts
Lift in a Recipe B:
If Recipe B shows a positive lift, this indicates that the ML algorithm driven personalized experience is performing better than the random allocation baseline.
Underperforming banners:
You can check the reporting metrics by segments to see if certain banners show more favorable results for specific profiles.
when to use Auto target:
In documentation, there is good examples and also mentioned about when you should choose Auto target - see the below link
And auto target faq help too. https://experienceleague.adobe.com/en/docs/target/using/activities/auto-target/auto-target-troublesh...
Adding @kandersen for more help in this thread.
Hope this helps.
Hi @vivin1 Adding comments based on my understanding for above scenario.
What is my control version now? Is it banner 1 or a random allocation of banner 1, 2 & 3?
In Auto-Target, there’s no true “static” control like a traditional A/B test. If you decide a control as banner 1, then Auto-Target may treat it as the "baseline." However, more commonly, Adobe Auto-Target will compare results based on personalization performance instead of a single fixed banner. This means it might consider a randomized version (random allocation of banners 1, 2, and 3) as the baseline.
What is my Recpie B: An auto-targeted personalized allocation of banners1,2 & 3?
Recipe B would dynamically assign either banner 1, 2, or 3 to individual users based on how the algorithm model predicts
Lift in a Recipe B:
If Recipe B shows a positive lift, this indicates that the ML algorithm driven personalized experience is performing better than the random allocation baseline.
Underperforming banners:
You can check the reporting metrics by segments to see if certain banners show more favorable results for specific profiles.
when to use Auto target:
In documentation, there is good examples and also mentioned about when you should choose Auto target - see the below link
And auto target faq help too. https://experienceleague.adobe.com/en/docs/target/using/activities/auto-target/auto-target-troublesh...
Adding @kandersen for more help in this thread.
Hope this helps.
Views
Like
Replies
Views
Likes
Replies
Views
Like
Replies