Would A/B Testing be still important for SEO?
If you log in to almost any business web analytics account and look at what sources the site receives most of your visitors, you will always find organic search traffic as the largest or second largest source.
And the finer it is, as organic traffic does not have to pay for any gimmicks or impressions – it’s all earned traffic. Therefore, SEO traffic is often very cost-effective for a company.
But SEO as a source of traffic significantly differs from paid media in that it cannot be tested.Or is it said that the SEO traffic could not be tested, but modern software and more advanced content management systems make it possible to test this very important channel.This blog is based on the initialisation of the SEO Marketing A / B test at the Search Marketing Finland event. You can see the materials in this presentation embedded below:
Why should SEO be tested for A / B?
We really have a great deal of responsibility for the SEO agents, because in general we control the web development of either our customers or our organisation, and we create a strategy to find the sites as broad as possible in the relevant terms of organic search.
One of the challenges of doing this is that we often have to rely on best practice and our own tests and we can never say for sure what the effect of a particular action is in the context of that particular site. This is what makes Google’s heavier machine-learning algorithms, which control the search results and the weight of different signals on the fly countless times a year.
In other words, we often make recommendations whose effects we cannot be sure of. Since most of the implementation of SEO measures requires either technical development or content refinement, the recommendations and their implementation have a real cost.And sometimes the best practice leaning or your own experience guides you in the wrong direction. Here’s an example of a Distilled office test where best practice implementation reduced organic visibility by 8%
Differences in SEO A / B Testing and Traditional A / B Testing
Generally, when talking about A / B testing, for example, in terms of online advertising or conversion optimisation , testing is done at the user level. This means that if there are even 10,000 people in the test, typically 5,000 people will see one version of the page and another 5,000 will see a different version of the same page:
A / B testing at user level
Conversion Optimisation A / B Testing Principle: Different versions for users are displayed on the same page.
SEO’s A / B testing works on a different principle, as it divides the side set into two parts, for example, leaving half of the pages in their original form and implementing the change in the other half. For example, an online store can have 400 product category pages, of which only 200 are changed. Thus, there are 2 different product category pages (original and modified) that are exactly the same for all users.
The principle of SEO A / B testing
SEO Principle of A / B Testing: All users see the same pages, but not all pages of the same type are similar.
Why then do the pages not be tested at the page level but are tested at the page type level? It is simply because you cannot divide one page in two and display all the page viewers, including the robots, exactly the same version of the page. Two may not be one. Thus, pages of the same type, such as product category pages, must be divided into two or more groups, of which only part is subject to changes.
In this case, the site will still have unmodified pages as a control group to which the performance of the modified pages can be compared.
It is important to remember that search engine bots must be provided with the same versions of pages as users, otherwise we will go to cloaking . Cloaking is strongly opposed to, for example, Google’s Webmaster Guidelines , and can easily lead to a drop in your site’s search results.
What is technically required for SEO A / B testing?
Technically, A / B testing for SEO can be done in a few ways:
If the publishing system allows it, the changes can be implemented directly there in bulk. Of course, it is also possible to make changes manually at the side level, but it is very laborious and therefore not efficient in using resources.
Alternatively, testing is done outside your own publishing system in another software that changes the content created by the publishing system before it is served to users and robots.
The image below shows the Distilled Optimization Delivery Network (ODN) policy:
- SEO stack for SEO A / B testing
- An example of a test stack
- In practice, using ODN, the process goes as follows:
The user or robot wants to see the page of the site and sends a GET request to the server. This happens whenever you try to load a page.
The server receives the invitation and sends the page to the source requester in the format reported by the publishing system.
At this point, ODN comes and makes a change to the source code of the page. That is, ODN is in order after CMS, but before the visitor’s browser.
If a site uses a CDN (Content Delivery Network, such as Cloudflare), the ODN will send the modified page source CDN.
The CDN sends the page forward to the party requesting the page.
Thus, the page the visitor requested may be different in content from what it appeared in the publishing system.
The thing sounds and looks complicated and it is technically the same, but the changes happen very fast and the whole process is usually carried out in tenths of seconds.
Test items and requirements
What pages can you test then? Actually, almost all the pages with more than one page. In practice, you can test product category pages, product pages, site pages, or even blog articles.
A / B Testing Items
A clear page that cannot be tested by A / B is the home page of the site because .. well, one page does not get divided in any way.
At practical level, the following things can be tested, for example:
- Metadata for pages
- Page titles
- Adding or deleting site elements (for example, links in the sidebar to previous blog posts).
- Adding Rich Data to Page Templates
- The number of products displayed on the product listing page
Summary
SEO testing is still very challenging at the moment due to the amount of data and technical complexity it requires. However, I firmly believe that in the future, those companies that are open-minded to do a variety of SEO tests will be most successful in searching, whether it is A / B testing or other systematic testing of how search engines respond to changes on the site.
In the best case, the site can systematically experiment with big and small changes and know their impact before knocking the technical team or content providers.
[vc_row][vc_column][ut_header title_linebreak_mobile=”on” lead_linebreak_mobile=”on” font_size=”26″ title=”Latest from the blog”][/ut_header][/vc_column][/vc_row][vc_row][vc_column][ut_grid_blog numberposts=”6″][/vc_column][/vc_row]