Protect SEO Rankings When AB or Multivariate Testing

Posted on 09. Feb, 2012 by in Digital Marketing



Don’t forget where your audience comes from when planning an AB or Multivariate test. Many tests that focus solely on improvements to user engagement and conversion do so at the expense of on-page optimization. This ultimately impacts rankings for top search terms.

AB testing example

Ensure your AB tests don't lead to indexing issues

Common Concerns and Myths

This simple factor is often overlooked amidst the common concerns (and myths) about the possible SEO impacts of AB testing or Multivariate testing. These include: duplicate content, cloaking, page load speed and lost link equity.

Cloaking isn’t an issue when using an optimisation tool as most of these will use Javascript to present content variations, which is not executed by search bots. Search engines will only see the original page and content for the duration of your experiment.

The other concerns focus on the nuts and bolts of testing and are easily remedied with better knowledge of search engines and indexing options for alternative pages:

  • Use of the rel=”canonical” tag to nominate the version of a page you want indexed
  • Use of the robots.txt file to disallow crawling for alternative URLs
  • Permanent (301) redirection of alternative page URLs to the original page URL once the experiment is complete and top performing version has been chosen and implemented

Variation pages can only be reached if there are links pointing to it. The measures above should keep the variations from being crawled and indexed.

Never hide the original page from visitors. If a non-original variant is run for 100% of visits for more than a month this will lead to indexing issues. Always show the original page, which is visible to search engines, to a reasonable percentage of users.

The Forgotten Element

If the page you are testing is potentially a landing page for search visitors, the main issue is that you don’t optimize for conversions by sacrificing traffic. 15% of not very much is less helpful than 12% of a lot.

By following the steps above, search engines will only see the existing (or ‘control’) page content while your AB experiment or Multivariate experiment is running and will not be affected by the test.

To prevent the results of an experiment adversely impacting SEO once implemented:

  1. Do your homework before designing the alternative copy, images, video and graphics. Fully optimized content will follow best practice in writing for the web, accessibility, search optimization, usability and creative design.
  2. Only test alternatives that you are confident are also optimized for search e.g. ensure there is sufficient accessible text to allow search engines and assistive reading devices to determine the focus/relevance of the page. If your original page is relevant and keyword rich but your variants are not this will lead to indexing issues.
  3. Your experiment should end when sufficient data has been collected. If you have few visits it will take more time to collect data so remember to keep your original page updated and fresh.
Multivariate Testing Example

Test variants that are optimized for search

Google Website Optimizer offers a wealth of helpful hints on testing best practice and website optimization.

Your end result should be a page that out-performs the original in increasing both traffic and conversion. To do this your digital marketing, content, analytics and user experience team must work together. We call this approach ‘conversion optimization’.