The Impact of A/B or Multivariate Testing on SEO

Chapters:
Article written by Gabriel
Community manager at PadiCode
Google+ profile

This is a guest post and the views of the author do not necessarily reflect the views and opinions of padicode.com. The author, Traian Neacsu, is the co-founder and Director of Search for Pitstop Media Inc, a top rated Vancouver SEO company.

Unless you’ve been living under a rock for the last few years, you have heard of A/B or multivariate testing for websites, aka conversion rate optimization. 2011 being coined by many “The Year of the Conversion Rate Optimization” and the fact that you’re reading this blog makes me think that you already familiar with jargon like CRO, LPO, MVT or A/B/. And, if you a smart marketer or website owner, then you already tried and reaped the benefits of website testing for conversion optimization.

Figure 1 – Image credit

Unfortunately, it is very common for the information-overloaded, super-busy, over-worked, {insert your other energy draining power here} online marketer not to be aware of the implications of website testing on search engine rankings. If your tests go live on a production server without some minimal SEO check up, you’re swimming in shallow waters. However, a well versed CRO consultant should make you aware of the possible risks of testing and your SEO company should provide some of the best practices to help you with the optimization process. Or, read below.

How Does Website Testing Affects SEO?

Here are some risks one has to take into consideration when testing and optimizing for conversions.

Duplicate Content Issues

When you setup an A/B test, you will create 2 URLs, i.e. www.site.com (aka the control) and www.site.com/index1.php (aka the variation). If you’re changing only some images and a couple of headlines on the variation page, you will end up with two highly identical pages, and this poses duplicate content issues. This might come as a surprise to some, since A/B testing is implemented with JavaScript redirects, which are supposed to be safe. But JavaScript is not to be blamed here.

Those pesky variation URLs will be indexed if somehow Google is able to find about them. Say you shared the test URL on a Google Webmaster Forum to clarify some technical issues, or you’ve mentioned it on your blog or someone came to your website, was served the test URL and then he published a link to your page, on his own website. To extreme cases, Google will use its own Google Toolbar to find out about new URLs (surprised about the Toolbar indexing? Then, please read this patent from Google). In all these instances Google will crawl and index the test URLs, and voila, you’re good for dupe issues.

Figure 2 – You don’t want your competitors to see these testing ideas, isn’t it?

With multivariate testing things can go even worse. Since multivariate testing doesn’t redirect to other URLs and keeps a single URL for all possible combination of the test, that can lead to hundreds of similar content URLs, which again turns into duplicate content problems.

Lost Backlink Equity

A little bit earlier I mentioned about people who visit your site and are served the test variation URLs of A/B tests. If they like the landing pages they will tweet, socially bookmark, or blog about them, thus creating backlinks to your site. And backlinks are generally good for rankings, right?

Once you’ve identified a test winner and stopped the test, the variation URLs will stop existing also (given that the winner is not the control URL). But those backlinks from other sites will still be alive. Google will pass Page Rank (if the backlinks are followed) to the landing pages and from there throughout your website only as long as the variation URLs resolve to 200 OK response headers, or at least 301’s. The Page Rank and your search rankings might be affected if the test is improperly concluded.

Additionally, if someone clicks on those links and your pages are not there anymore then they will get a 404 Page Not Found Error, which is a bad user experience no matter how great the custom 404 page is.

SERP Bounce Rate

Search Engine Results bounce rate is not the same with the website bounce rate. A SERP bounce is when you bounce back to the Google search results page, without visiting several pages or spending a certain amount of time on the URL you clicked on.

Google has a feature called Block all www.site.com results, which is available when you click on a page in the SERPs and then bounce. Keep in mind that SERP Bounce is also a very strong user feedback signal to Google and also an SEO factor.

Figure 3 – SERP bounce in Google

When testing, many marketers measure the success of the test by looking at macro conversions, which can be a lead sign up or a purchase or a subscription. More advanced online marketers look and optimize for micro conversions and even event conversion (no page views). But not many take SERP bounce rate into account. If the number of visitors who bounce from Google increases, you jeopardize your rankings.

An increase in the SERP bounce rate can signal Google that the page is not worthy anymore and its rankings can decrease. Unfortunately, the SERP bounce rate is a bit more difficult to measure and integrate with your test, it requires some additional manual work and that makes many of us ignore this decision making metric.

Let’s finish the risks list with a couple of factors you don’t need to worry too much (at last )

Cloaking

While technically speaking showing different content to visitors than to Googlebot is considered cloaking, website testing in most cases is not considered cloaking. Just follow the official guidelines from Google and you should be safe.

Page Speed

While not a significant contributor to search rankings, load time is still an official SEO factor, affecting about 1% of the search queries. Unless your tests generate really slow pages (say more than 5 seconds for your content to start displaying in the browser) page speed should be the last of your concerns.

What Should You Do?

Below is a list of best practices that you can follow to make sure you’re safe with SEO when testing.

1. Add the following meta tags in ALL variation URLs:
<meta name=”robots” content=”noindex,nofollow,noarchive”>
2. Put the test pages in a dedicated directory (i.e. mysite.com/testing/) and exclude the directory with robots.txt.
3. In your variation URLs use the rel=”canonical” meta tag to suggest to search engines the canonical URL that they should index.
4. Avoid linking to your testing directory/pages from anywhere.
5. Do a permanent redirect (301) for all tested URLs to the winning URL once the test is concluded.
6. Test only with a portion of the traffic, if possible. This is to be safe with the SERP bounce rate.
7. Watch the rankings for the page you’re testing. That will be very difficult for template based MVT testing, but easier for A/B testing.

Happy testing and please leave a comment!

The Impact of A/B or Multivariate Testing on SEO by

1.100.000+ email leads collected with

PadiAct helps you get more email leads. More and better.

FIND OUT MORE