Take your SEO to the next level with SEO testing

Aleyda says: "I believe we have reached a point in SEO maturity, where we can stop focusing so much on avoiding SEO issues and errors or lifting credibility and executability of our websites. Instead, we can increase our focus on how and where we can build up to grow our results. Let's take our results to the next level and become much more competitive through developing and building new initiatives. There are lots of opportunities to connect to our search audience much better and using the many new SEO testing platforms will become much more of a norm. They are easier to use and cheaper so, if you haven't started using these testing platforms yet, I highly recommend you start doing so right away."

What SEO testing are you running most often? Are we talking about split-testing or something else?

"A/B testing, doing an experiment or developing events. Sometimes we are asked to prove potential returns from certain efforts to the decision-makers. This could require non-trivial efforts involving technical, content, and even product changes. I believe that we now have tools to help us that can be easily implemented through JavaScript. For example, Semrush and SEOTesting.com, which is another tool that can easily connect with Google Search Console. Then we have all the tools like Ranksense that allow CDNs to leverage CloudFlare Workers to run these tests.

Our testing can become pilot projects for bigger efforts and helps us show their impact without having to wait for ages. Furthermore, they ensure that we have good testing frameworks and establish control groups to develop these experiments in a way that actually makes sense. This means we can check and validate that something makes sense in our own specific context without implementing millions of URLs. We can remove a lot of uncertainty and move beyond generic best practice recommendations that may or may not work for us."

There are so many different things that can be tested. How would you know which thing to test first?

"There is a lot of low-hanging fruit, such as the metadata. In fact, a recent Google update has lowered our control there. A timely example is tweaking the titles and meta descriptions that we show in search results, to make sure Google picks up the actual content of title tags.

Next, there's a lot of conversations about whether more content is always better. We need to understand that it depends. What is the nature of the page? Is it a blog post or an article that only has an informational intent, or is it a category or product page? You also need to consider the role of that page in the customer journey.

Then, there are other questions such as, 'How many products should I be showing to drive traffic to product pages? How many internal links?' or, 'How can I optimise navigation to refer more traffic to the most important pages?' Testing allows us to validate a lot of best practices that we know, generally, should be good for SEO - but are often not trivial to implement. We can confirm that initiatives will actually be good for us and won't waste significant effort."

Should an SEO be concerned about the version of the test that Google is seeing? Several years ago, Google struggled with JavaScript and now it can see within JavaScript and the content that comes from there. What version of the test is Google likely to see, and is this likely to impact the SEO and the conducting of the test?

"Yes, it is a positive development that it is now able to render and process JavaScript. It's the same with CloudFlare Workers, which we can leverage to show the right version, or the tweaked the version, of the pages that we want to test directly through the CDN. This allows us to see the variations we want and learn which ones are performing better. Then, if our hypotheses don't play out as expected and we decrease in rankings or traffic - it should be ok. The idea is that we confirm if we are right or not before we release all the changes across all of the millions of URLs on our sites."

How do you select a winner? How much data is required and what kind of percentage confidence does the software need to declare a winner?

"Of course, this is not trivial, and it will vary. But you need certainty from any testing, and this framework will allow you to establish this. You need to be informed about the level of traffic, and length of time required, to run this test and declare a clear winner. Many of the tools and platforms I've mentioned allow you to establish and configure the whole testing environment, so your experiments make sense and provide you with enough confidence before selecting a winner."

Typically, are you looking at a certain number of visitors or a certain amount of time to actually make a test valid?

"Yes. If you have a website with only a few hundred visits, it will be very hard to test. You should be looking to have minimum levels of traffic. I think Semrush's SplitSignal recommends dozens, or hundreds, of thousands of visitors."

You've mentioned a couple of bits of software, such as SplitSignal by Semrush and also SEOTesting. Which tool is best to get started with? And what are the pros and cons of each tool?

"SEOTesting.com integrates directly with your Google Search Console and allows you to run time-based SEO tests as well as A/B tests. This is a good starting place if you don't have the technical capacity to implement something like JavaScript to display the changes and variations across different groups of pages. It allows you to run different sorts of tests and integrates Google Search Console.

Then you have tools, such as Semrush's SplitSignal, that use and leverage JavaScript and Google Tag Manager. Chris Green recently presented at BrightonSEO, demonstrating how you can develop your own testing framework using Google Tag Manager - so you can run this completely yourself if you want. The main benefit of tools like SplitSignals is that they provide you with a very nice interface to analyse, visualise, and easily establish how your tests are performing. However, implementing the test yourself is completely doable.

Finally, there are platforms like Ranksense. They allow you to implement testing and will run directly through the CDN using CloudFlare Workers, so you won't depend on JavaScript."

For running split-tests, I use Google Website Optimizer. Are you familiar with this tool, and is it an option for SEO as well?

"No, because these are UX-oriented. When UX people run this test, there are certain best practices we need to take so they don't get confused with the different variations - they don't end up indexing twice, or cause inconsistencies with the way we want to see our website configured. We usually canonicalize one version to the other or redirect from one version to the other. In general, these tools are designed to run this test at the user level. These are tests that are designed to test and validate how it will affect our actual rankings traffic - from an SEO standpoint."

What should an SEO stop doing to spend more time on SEO testing?

"There are a lot of rules of thumb that we have taken as a given. For example, that you need to optimise your title and meta description to include terms, and you can use certain characteristics and variations. But how much is too much? We've now seen from a recent Google update that sometimes the right title is not shown because we were over-optimising the title and it was too long.

Therefore, the title might be one of those elements that has a higher impact because it affects relevance, click-through rate, and engagement in SERPs. We can test how we are wording, or showing, information regarding the number of products that we feature, or information regarding the characteristics of certain types of pages. It's very straightforward to do a little bit more testing of our titles.

Also, effort that is non-trivial. Sometimes we recommend activities that we know are beneficial, but they are harder to implement because of their resource requirements. For example, adding certain types of descriptions on our category pages. We have millions of them - so how can we write unique descriptions for them? On the one hand, we now have tools to try to automatise this process. On the other hand, even automating a lot of this process takes lots of human effort and copywriting resources.

How do you make sure the impact will be positive from an SEO, and traffic standpoint? You can run a test with a group of category pages. This will allow you to see if the impact is positive and tell you if it makes sense to add content, and help you understand the type of content to add to those particular pages."

Aleyda Solis is an SEO Consultant and Founder at Orainti.com.