a/b redirect test an entire website




Has anyone had any recent experiences, or advice to offer for configuring a multi-page redirect test between two sub-domains?

The concept is fairly simple – but I am anticipating some complications in achieving a robust test.

It could be thought of as a/b testing a new version of an entire eCommerce website   from www.example.com to ww2.example.com 

One complexity is that some of the URL structure on the new website is not the same as the original so there are two types of redirects needed:

a) Fixed pages

  1. eg.       www.example.com/homepage    

      to ww2.example.com/homepage

b) Dynamic pages with a different URL   🙂

  1. e.g.      www.example.com/electronics/televisions/sony

     to ww2.example.com/electronics/entertainment/televisions/sony

As with most target activities there will be are a number of ways we could approach the setup, so looking for any words of wisdom, areas that we should pay particular attention to, or could get caught up in.

We must ensure that:

a) We have clearly defined and consistent A and B audiences.  I'm hopeful that standard profile scripts will work fine here.

b) We must ensure a consistent experience on returning visits. So after being bucketed as a B user, entering back to a page on either subdomain will take them to the subdomain / page appropriate to that audience.

c) The redirects must work both ways   (So a ww2 user always gets ww2 pages, and a www user always gets www pages)

d) Many other pages on the site are common, so we don’t mind whether people see the www or ww2 version, as long as we track this appropriately in Adobe Analytics we should be fine….

f) I anticipate this being an A4T integration so we can evaluate the overall user journey between the sites, and between visitor sessions and drill into segmentation.

g) Our success at a Macro level will be overall sales conversion and we need to achieve a robust view of this between the A and B groups.

I appreciate this post covers a lot of ground, and could raise lots of different questions - which i could tackle separately - but in the first instance just looking for your experiences, any do’s and don’ts for best practice in this situation, or any warning of “don’t try and attempt this!” from the community 🙂







Some initial thoughts:

  • I think you are on the right track with using a profile script to segment out the audiences in a way that will be sticky
  • You could run 2 global redirect campaigns (one for each group) that uses a js offer rather than an out of the box target redirect offer so that you have control over what it is doing. Example code below.
  • You would need to make sure to pass the mboxSession to the new domain so that conversions work appropriately.
  • I would keep all the pathname matching to the apache/nginx/httpd.conf side rather than trying to tell Adobe all about it, unless it is only a few differences.
  • If you were able to relax any of the requirements, it would make the test easier. For instance, group B gets globally redirected to the top of the new store regardless of entry.

A rough js client side redirect for just switching subdomains:

// www.test.com/store/tshirt/

// www.test.com

var host = window.location.host;

// store/tshirt/

var pathname = window.location.pathname;

// the new subdomain to go to

var newSubDomain = "ww2";

// split on . and take the first result

var hostArray = host.split('.');

// build the redirect

var newLocation = "https://" + newSubDomain + "." + hostArray[1] + "."  + hostArray[2] + pathname;


Will continue to think on it and comment if something else comes to mind.




Thanks for the detailed response Eric.  Feels like we are on the right lines.

A bespoke redirect in JS would give us more flexibility, although I thought from what i'd read is that only the out of the box redirects keep the data in synch between both the Mbox and Analytics calls firing correctly at different positions on the page the redirect happening as part.

All our previous redirect tests with custom JS have produced messy conversion data in an A4T activity,

I presume this is what you mean about passing the Mbox session to the destination page.




I don't know a whole lot about A4T but I would venture to say that it is just looking for the right tokens in the redirect. Something like this should at least help prove out the point (along with adding session if bouncing between two different domains):


The session piece has more to do with the cookie pool. abc.com and def.com don't get to share the mbox cookie which contains the session. Hence passing mboxSession will overwrite the generated session on the new domain.




Thanks Eric,

The A4T considerations were raised from here- Redirect offers - A4T FAQ   but will give your method a try and see what happens in the data.

We manage mbox cookies at a domain level, so i'm hoping the as we are only changing subdomains, these should persist as if it is a single website.

e.g.   www.test.com  Vs ww2.test.com