Backlinks pointing to the B page of an A/B test.

[JoeNYC]

To rel-canonical or to 301, that is the question.

We're frequently running an A/B split test on our home page to optimize conversion.  As a result about 10,000 backlinks to our homepage point to the B page.  (If we're running a test when a blog or newspaper checks us out, there's a 50% chance they're diverted to the B page. So when they copy our home page URL, they're unknowingly copying the B page link.)

We can't contact all of these sites and ask for them to change their links.  A lot of the links are from big organizations that aren't interested in tweaking the links of old articles.

So should we rel-canonical or 301 the B page?  We consistently use the same URL for our B page tests, so we'd only have to 'fix' one page.

Thanks in advance!

[JoeNYC]

Thanks Keri.  Appreciate the offer, but I'm all set.  We just 301d the page after the test was done and have decided that we'll only multivariate the home page for now.  We've now had 3 major press links point to the testing version of our home page, so we've decided our best bet is just to set up all tests as multivariate so we don't create two different URLs.

Thanks again.

[KeriMorgret]

Hi Joe,

I'm following up on older, unanswered posts. Did you find a solution, or are you still looking for advice?

[JoeNYC]

Thanks for your thoughts.

While a canonical shouldn't impact a split test, a 301 would create a loop and make an A/B test impossible.  Although, if we were to 301 this page, we wouldn't use that same page for testing.

As far as multivariate vs A/B.  The design changes are often significant enough where a multivariate is impossible.

FYI - we use Google Website Optimizer for our tests.

Thanks again.  Appreciate your weighing in.

[IPINGlobal54]

From a logical point of view (or thinking out loud as it were) if you are consistently using the same B page, that in itself will be where the problem is likely to be.

For example - if you canonical the B page, and continue to run the B page anyway, I would think you would end up creating a loop, or at minimum a strange redirect because the script in you A page for the split test to function will/could point it back at page B

If you 301, I would think the same could apply, and still not solve the issue.

If you ran the tests using the multivariant test format instead, I would think this is less likely to cause issues because you are then just switching segments of content on your landing page, rather than urls, therefore eliminating the problem altogether.