Cross-Domain Landing Page Split Testing

Follow
Spencer Skeen
  • Edited

Good afternoon,

We are currently looking into testing several static landing pages and we are having trouble integrating with split.

One of our landing pages would exist at a sub domain, i.e. "promo.domain.com" and redirect to the root domain for the consecutive pages. The other landing page will be at our root domain, i.e "domain.com" which the consecutive pages will be on that root domain. Is there a way to track a split test experiment for this type of set up?

We've gone through the docs and don't see anything that would help this type of set up. We believe it would work if we could programmatically set the treatment value (on/off) on the consecutive pages based on the landing page they originated from?

Hopefully this make sense, let me know if I need to clarify anything.

 

Thanks in advanced,

Spencer

0

Comments

8 comments

Date Votes
  • Bilal Al-Shahwany

    Hi Spencer,

    Its possible to implement Split Javascript SDK on each static page as needed, as long as you are using the same Split API Key, you can calculate the treatment for any split (or experiment) that is added to the environment associated with the API Key.

    Are you seeing any issues implementing the JS SDK?

    Thanks

    Bilal

    0
    Comment actions Permalink
  • Spencer Skeen

    We are currently running split successfully on our root domain (domain.com) which is an Ecommerce store. We send events to Split via the Segment integration on "AddToCart", "Purchases" and several others and has been working great for feature flagging. 

    As for our landing page testing, it is set up differently. The landing page we are looking to split test is located at "promo.domain.com", users will be shown a different homepage variation which will then redirect to our root domain where they can shop our products and perform the events (addtocart, purchase). No events take place on the promo.domain.com landing page. All events take place at the root domain. Now is there a way for these events on the root domain to know the user came from promo.domain.com and not our root domain homepage. 

    Hope this make sense. 

    0
    Comment actions Permalink
  • Bilal Al-Shahwany

    Hi Spencer, Yes. From Split point of view, the events created in any page can be correlated to any split, as long as you are tagging the events and impressions with the same user id (same traffic type) and same environment.

    It does not make a difference where the impression is generated from, as long as the SDK is using the same API Key in all pages.

    Thanks

    Bilal

    0
    Comment actions Permalink
  • Spencer Skeen
    • Edited

    Okay that makes sense, thanks.

    Now let's say I wanted the promo.domain.com landing page to be the 'on' treatment for experiment "LandingPage_Test" and the domain.com landing page to be the 'off' treatment for given experiment, what would this implementation look like?

    0
    Comment actions Permalink
  • Spencer Skeen
    • Edited

    Or would it be better to have two experiments, "LandingPage_Test_Subdomain" and "LandingPage_Test_Domain"?

    Although, correct me if im wrong, you cant compare metrics across two different experiments?

    0
    Comment actions Permalink
  • Bilal Al-Shahwany

    Hi Spencer, it makes more sense to have one experiment "LandingPage" with two treatments, "on" for promo.domain.com  and "off" for domain.com, this way you can compare both on and off populations in the same metric and determine which landing page show positive results.

    The split would be called on the previous page (like login), if there is no previous page, then the split is called in an initial page to determine which landing page to route to.

    Thanks

    Bilal

    0
    Comment actions Permalink
  • Spencer Skeen

    I made a new post about this but it's related to this set up. We are currently serving a treatment to our anonymous users based on an attribute we provide to split. When we attempt to compare metrics, there is a message, "Statistical Comparison not possible because metric is grouped across anonymous". What exactly does this mean, and is it possible it fix?

    0
    Comment actions Permalink
  • Dave Murphy

    Hi Spencer, I responded on the other post.  You need to change your metrics from 'across' to 'per'.  I included some additional detail there, along with a link to a video.

    dave

    0
    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post