A/B Testing SEO friendly

A/B Testing for SEO

I love testing. In my opinion, Website Testing is one of the most important conversion optimization techniques. However, it should be used taking into consideration the wider context of online marketing. In this post I will show a case study on how to account for SEO when doing A/B testing in websites. I will first discuss how A/B test can hurt SEO and then provide a way to solve the issues that might be caused by A/B tests.

Is My AB Test Hurting My SEO Efforts?

We used an A/B testing tool to conduct the experiment, and tested several homepage versions under the following URLs: example.com, example.com/index2.html, example.com/index3.html, etc. Two weeks after initializing the test, in a routine check of the website's indexed pages at Google, using a standard search query (site:www.example.com) we found, to our surprise, that the alternative homepage version had been indexed by Google.

This is so surprising because, theoretically, A/B tests redirect using JavaScript and they should keep the variations uncrawled and unindexed. This certainly holds true to multivariate tests, which do not redirect visitors from page to page, but show different versions inside the same page. But in this article we will show that it does not always hold true when it comes to A/B tests (read more about the differences between AB and Multivariate tests).

Website Tests Do Not Occur In a Void

To identify the source of crawlers into the alternative homepage, we checked for backward links to that page using the Google operator link (e.g. link:www.example.com/index2.html). There was a single external link pointing to that page. The link to the alternative URL appeared in an article quoting the website published in a business review website. We realized that the article's author, attempting to insert a link, entered the website's URL during the trial period and was served an alternative homepage version. He then copied the alternative URL and pasted it in his article. The crawlers' path to accessing and indexing the alternative homepage version was clear. Mystery solved!

There is an important lesson to be learned from this case: although A/B tests work under a controlled test environment, one that is also blocked from search engine crawlers, tests do not occur in a void. Test participants (i.e. site visitors) operate as the gate out of the closed test. Participants are unknowingly given test versions, which they can save as bookmarks, or add as a link in their blogs, unaware they are linking to a test version.

Accounting For SEO When AB Testing

Our solution involved two steps:

  1. We immediately added the alternative URL to the site's robots.txt file to disallow crawling. This method guaranteed that the alternative homepage version would not be indexed by search engines.
  2. After the A/B testing was complete, the chosen page version was set to the original homepage URL. We then added a permanent redirection (301) of the alternative URL (/index2.html) to the original homepage URL. This would channel any traffic attempting to enter the site via alternative link versions, correct the homepage, and avoid dead end traffic caused by broken links.

This example shows us that, once again, search marketing should be viewed as a conglomerate of variables, where SEO, paid campaigns and Web Analytics are intertwined. Each effort to optimize a website should strengthen all others to increase synergy and boost revenues.

Learn more about SEO Analytics

Subscribe To Our Newsletter For Monthly Updates



Your e-mail will be kept private

Bfri | February 2011

Adding an URL to the robots.txt does prevent Google from crawling the page. But it will still be indexed if another site points a link to the URL. You should set the robots meta tag to "noindex,follow" instead.

Fer Vazquez | February 2011

I came across a similar situation. Definitely agree with a holistic approach to web testing where marketers, developers, designers and analysts are all actively involved both during test design and the results evaluation.

About the SEO bits... like Bfri said, if you can modify your HTML easily, then the meta robots is a better way forward. I'd also add a wee canonical tag pointing to the original homepage.

Chris Goward | February 2011

Well written, Michal!

We've seen this happen quite often as well and always recommend the 301 redirect after competing an A/B/n test. There are several ways that CRO testing supports SEO efforts and SEO practitioners shouldn't be afraid of testing. As long as it's done right.

You may have already seen the tips for how to do CRO without messing up your SEO efforts on the WiderFunnel Conversion Optimization blog too:
http://www.widerfunnel.com/conversion-rate-optimization/how-to-combine-c...
http://www.widerfunnel.com/conversion-rate-optimization/how-to-combine-c...

It includes the same 301 redirect tip that you mention as well as other considerations.

Dave Morgan | April 2011

This is a helpful post, although would have been nice to know which tool you're using. I'm assuming it was Google Website Optimizer, which does these sorts of redirects for A/B tests?

It's worth pointing out that not all of the available testing tools work this way. Some of the tag-based tools like Autonomy Optimost let you do whole (well, almost whole) page A/B testing without redirects, and the non-tag tools like SiteSpect do as well. And the CMS tools like ATG and IBM WebSphere support A/B testing of whole pages without redirects.

Avoid the redirects, and you basically avoid this SEO problem whereby alternate page versions get inadvertently indexed by search engines.

Full disclosure, I work for SiteSpect, a provider of testing tools. But don't hold that against me. :) I think this info is still very helpful.

thanks!

Dave

Jori Ford | March 2012

Dave, I love that you gave full-disclosure. I did notice you pointed out CMS systems like ATG and IBM WebSphere. You're correct in their support of A/B testing, however it's in the same manner that Michal explains multivariate testing.

Anytime you use a tool that dynamically alters components on page, providing different variants of existing elements, in consideration of Michal's definition, I'd group it under multivariate.

So, what type of tool is SiteSpect and does it provide some semblance of these types of testing tools or is it more like Liveball?

BTW - Great post Michal!

After so many "demystified" and "debunked" articles, it's refreshing to see someone who looked at the bare bones basics of how a page gets indexed and didn't rely on the engine to not make any mistakes.

Michal Bitton Nassimian | April 2011

Thanks for all the valuable comments to this post!

Dave, thanks for presenting testing tools/products that seem to work in a different way which may bypass the need to perform redirects following an A/B test that serves various page versions under different URLs.

I should point out that the solutions presented in this post and in some of the comments, such as adding tag and using redirects after the test is complete require minimal efforts and are very simple to perform.

Thanks!

Michal

Iulian Grigorescu | April 2012

I had the same problem as Vishal. I think the problem is with 301 redirect.

Yesceeohhh | June 2012

You could have declared the pages "noindex, no follow" While A / B testing. If you have done that, you could have avoid google indexing your testing pages.
http://www.usjoomlaforce.com

Daniel Waisberg | August 2012

The Google Webmaster Central just released official recommendations regarding what to do when testing your pages: Website testing & Google search

francois | October 2012

hey,

I'm using prestashop and i'd like to run ab test.

I've design an alternate product page so I have :

product.tpl

and

product2.tpl

I need to assign a different url to reach the page with product2.tpl

How can i do this ?? (i dont know how to add paramater to the url)

Thanks

François

Marc | March 2013

Hello,
We've made an add-on that does exactly what you want : no code required. It serves randomly one of two templates and rewrite urls to track and compare results in Analytics : http://www.creation-sites-web.eu/add-on-prestashop-a-b-test/
Enjoy.

pradeep | December 2013

This article gives the light in which we can observe the reality. This is very nice one and gives in depth information.

Post new comment
The content of this field is kept private and will not be shown publicly.
Refresh Type the characters you see in this picture.
Type the characters you see in the picture; if you can't read them, submit the form and a new image will be generated. Not case sensitive.  Switch to audio verification.
Online Behavior © 2012