30 second summary:
- SEO A / B testing can help website owners understand whether the changes they are making to their website are having a positive effect on keyword rankings.
- Everything on a website can be A / B tested. However, the best variants to test are those that are directly related to Google's ranking algorithm.
- The best way for website owners to perform an SEO A / B test depends on the size of their website and their ability to easily undo changes.
- If an A / B test proves that a particular optimization is effective, a website owner can more safely make identical or similar changes to other pages on their website.
Sometimes it is difficult for SEO strategists to know which of the many changes we make to our websites will actually affect their overall SEO performance. Moving websites from page 10 to page 2 can usually be done by following SEO best practices, but getting to page 1? This requires a far closer look at the specific changes we make to our landing pages. Take SEO A / B testing, one of the best ways to narrow down and understand the effectiveness (or ineffectiveness) of certain optimizations.
Many digital marketers can use A / B testing functions in their PPC campaigns or analyze user behavior in Google Analytics. Fewer are using this powerful strategy to better understand which of their on-page optimizations are having the greatest impact on keyword ranking.
SEO A / B testing can be intimidating, but it's actually quite easy to do with the right tools. A / B testing not only helps SEOs identify the most effective optimizations, it also gives them a chance to quantify their efforts. Although SEO A / B testing is more commonly used by advanced SEO strategists, website owners who enjoy working in the backend of their website have a great opportunity to improve their SEO strategy through well-structured A / B testing.
What is SEO A / B test?
When it comes to controlled experiments, testing two variants is the basic building block from which all other tests are built. The A / B test only measures how a single variant affects a result. In the case of SEO, the results are either better, worse, or static keyword rankings.
It doesn't take much to get started. You need the Google Search Console for one (the absolute truth of your SEO rankings) and two well-defined variants that you want to test. Finally, website owners need to have a development or staging environment in which they can save a version of their website prior to SEO A / B testing in case the changes don't produce the desired results.
Best use cases for SEO A / B testing
Everything on our websites can be A / B tested. For SEO purposes, however, certain website elements are more likely to lead to improvements in keyword rankings due to the weight they have in the Google algorithm. Because of this, the following elements are the best use cases for SEO A / B testing.
Choosing title tags is very important and has a huge impact on search results. Changes to title tags are very effective from a ranking point of view as they directly affect the click-through rate (CTR). Google has a normalized expected CTR for search queries. If your landing pages continually drop below the mark, it will negatively affect your overall ranking chances.
For websites that already have a lot of keywords on page 1 and are therefore getting a lot of impressions, A / B test meta descriptions can be very useful. Like title tags, they have a direct impact on the click rate. An improvement can lead to significantly more clicks and thus to better rankings.
If you can, it's good to add schema markup to all of your web pages (you're missing out on any shots you don't take!), But if certain pages of your site still don't have schema.org markup, you can add one be good use case for an A / B test.
Internal links communicate with the architecture of the Google website and are also distributed Side rank on our websites. If internal links are displayed correctly, it can lead to dramatic improvements in keyword rankings, especially on larger websites with thousands of landing pages. Focus on links to headers and footers as these will greatly shift PageRank. For websites with product pages, you can use A / B testing to find the best anchor text for your internal links.
Adding good content to your landing pages is always beneficial as longer content implies current depth. Using a Landing Page Optimizer You can use this tool to improve the semantic richness of your content. With a subsequent A / B test, you can measure whether Google reacts positively to these quality signals.
Large groups of pages
For ecommerce websites or websites that may be adding a large group of pages at once, you can use A / B testing to measure whether those pages are being crawled and indexed in a way that will positively or negatively affect your existing rankings.
There are certain elements of the information architecture that are more specific to SEO. Google likes page experience features like jump links and carousels. Therefore, understanding the effects of adding these features to your web pages is another reason to do an SEO A / B test.
If you're making fundamental technical changes to your website, you can use A / B testing to measure how those changes affect keyword ranking. It also helps prevent your website from seeing any significant drop in rankings in the long run.
The types of SEO A / B tests
There are several different types of A / B tests that you can run on your website. This largely depends on the number of landing pages you have, as well as the category of variants you are testing.
A then B.
This type of test is the most basic form of A / B testing and simply compares the performance of a single page against another variation. This type of A / B testing is better for smaller sites and easier to implement, especially when you are sure the changes are going in the right direction.
This type of A / B test provides far more statistically significant results and can be run on large websites with hundreds to thousands of landing pages. Instead of measuring a single variant on a single page, choose two page groups of similar pages (I recommend at least 50 pages per cohort) and change the variant on all of those pages.
This common form of experimentation has the same core mechanisms as an A / B test, but increases the number of variants tested. Multivariate tests can be great for measuring user behavior. However, they are less effective at measuring search performance when you want to find out if a particular optimization has a direct impact on keyword ranking.
How to do an SEO A / B test
The easiest way to do A / B testing is to use site snapshots and rollbacks. Basically, you take a site snapshot, make the change to your site, and sit back for a week and watch what happens. If the website change has improved your SEO, you will see it in your benchmarking. If you made a mistake, just go back to a previous version of your website and continue with various tweaks.
Here is a simple step-by-step guide to an SEO A / B test:
- Take a site snapshot or perform a site backup Before you implement changes, you can revert to a previous version of your website if the change does not take effect
- Determine the variant to be tested (e.g. title tag, schema.org and the others) and make the change to a page or cohort of similar pages
- Wait 7-14 days to determine the impact of the change
- Compare the keyword rankings for the variant page (s) with the original. This can be done in the Google Search Console or in a Google Search Console Dashboard
If the tweaks have an impact, you can make similar changes to other pages on your site. More focused changes like adding keyword-rich title tags and meta descriptions don't necessarily translate directly to other pages. However, other technical optimizations such as schema.org and information architecture can be implemented across the website with greater certainty if an A / B test proves their effectiveness.
Some practical advice for SEO testing in hosting environments
How you proceed to step 1 depends on your developer and your staging environment. If you host on DigitalOcean you can take site snapshots. When working with WPEngine, you can choose a site backup to restore from. Try the best in class Version control from Git that allows you to revert to any version of your website.
Version control is similar to tracking changes in a Word document, but the history is never cleared. Even if you delete or change something using version control, there's a chronology of all the changes that were made – when a line of code was created, when it was edited – and you can roll back at any time.
It's also important to make sure that all of the pages in development and test lab pages on your website have the following Robot no index tag. Some SEO specialists might tell you that adding the pages to your blocklist in robots.txt or with rel canonical tags on the page would be sufficient, but at LinkGraph I've seen numerous examples of pages or subdomains being added to robots and showing up in search results for months. Adding a canonical tag is not enough to block the development domain from crawling and indexing.
The best strategy is not to give the robots an index. The even better strategy is to use both robots with no index and rel canonicals for extra protection.
How long does it take to know if the variant was effective?
How long you wait to measure your A / B tests depends on how often Google crawls your website. If you're testing parts of your domain that aren't crawled often, you may have to wait longer for Google to actually see your changes. If you change primary pages that Google crawls frequently, you can likely see if those tweaks had an impact within 7-14 days.
When done correctly, A / B testing can be a powerful way to gear your SEO strategy for maximum results. Not only can A / B testing help website owners make data-driven decisions, but it can also help SEO strategists prove the value of their work to clients or executives who may be wary of investing in SEO.
Manick Bhan is the founder and CTO of LinkGraph, an award-winning digital marketing and SEO agency that provides SEO, paid media and content marketing services. He is also the founder and CEO of SearchAtlas, a software suite of free SEO tools. You can find Manick on Twitter @ Madmanick.