AdWords is now Google Ads. Our new name reflects the full range of advertising options we offer across Search, Display, YouTube, and more. Learn more

Optimize
2.2K members online now
2.2K members online now
Discover how to select who is eligible to be in your experiment, and when they’re served experiment variations.
Guide Me
star_border
Reply

Traffic not distributed evenly among variations

Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

Hello - I am trying to figure out why my redirect experiment is not weighting the variants 50/50.  It has been running 13 days has sent 20 users to the original, but 63 users to the variant, which is not even close to a 50/50 split.

 

From the google analytics page: "By default, Analytics uses a multi-armed-bandit approach to manage the traffic to your variation pages, and to declare a winning variation. While there are clear advantages to the multi-armed-bandit approach (see below), you do have the option to distribute traffic evenly among your variations."

 

This would explain the weighting issue.  However, from the google optimize page: "Optimize currently distributes traffic evenly across all variations. We're working on adding in a multi-armed bandit weighting soon and like in Content Experiments, you'll be able to select which approach you want to use for each experiment. Watch the Optimize Release notes for updates on new features."

 

In addition, there is no toggle option when I create my optimize experiment, to set the weighting to distribute evenly.  I have to select 50% for each variant.  

 

Any help would be appreciated!!

 

 

1 Expert replyverified_user

Traffic not distributed evenly among variations

Google Employee
# 2
Google Employee

Hi Annie

 

Optimize doesn't use the multi-armed bandit at the moment (it was an option for the legacy GA content experiments) and distributes traffic based on the weights. By default (without you having to change the weights), it distributes traffic equally among variants.

 

Are you seen a difference on number of users or sessions? Optimize distributes new users according to weights, but as soon as a user is assigned to a variant, they stay in that for the duration of the experiment.

 

Could it be that some of your very active users are in your variant (perhaps your variant is doing a lot better in engagement) and this is why the number of sessions there is a lot higher?

 

If this is not the case, then could you share your Optimize container ID to have a look?

Traffic not distributed evenly among variations

[ Edited ]
Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭

Thanks for the response!  How do I check the number of users (I've only seen number of experiment sessions)?  Would I take the percentage of new sessions multiplied by the total sessions for each day, or is there an easier way to get the number of unique users?

 

Also, can you send me an email at annie at lenda dot com, so I can send my container Id?

Traffic not distributed evenly among variations

Google Employee
# 4
Google Employee

Hi

 

Yes multiplying the percentage of new sessions by the total sessions, would give you an idea. Another option that I like is making a segment for each variant so I can see what happens across all GA reports (i.e. active visitors). Make sure that you use Experiment Name and Variant conditions as a "sequence" in the segment builder.

 

If you still see a problem, you may use the feedback option from the top-bar overflow menu, to provide your container id and site URL.