Forum Moderators: open
We're currently looking into trying A/B testing on one of our sites. We basically want to randomize what is served to the user to compare results between two different versions of our site. Example: User lands on our page and sees Version 1 with red background color (root folder), while another user would see Version 2 with orange background color (sub-folder). Assigned landing would be completely random.
Since we will have two exact pages on the server, we must have a bullet-proof method of making sure we don't get ourselves hammered by a duplicate content or cloaking penalty.
Could anyone provide any help or suggestions before we start working on the project?
(We're running on IIS)
What you need to do is seperate out the bots, so they get a seemless experience eg, "cloak" the content as it was previous for the bots so they can get the same experience they've always had, and then when a user is identified, push them into the A/B test.
Depending on your site's architecture, setting up the "cloak" will be anything from easy to difficult.
I've done a few A/B splits myself, and the method I employed was to cloak the site so the search engine bots got the same old version, and then the user got the split run test, that way you also get accurate stats as only people participate in the split run test.
FYI, there is NO such thing as a "cloaking" penalty...;)
Cheers.
Let us know how it works out! There is scant info on the 'net about proper A/B testing, and it's one of the great tools any marketer has at their disposal to up ROI from search.
