6 SEO Split Testing Best Practices to Validate Your ROI

Brian Moseley

Feb 16, 20226 min read
6 SEO Split Testing Best Practices to Validate Your ROI

Site owners expect a return on investment (ROI) from their SEO budget, whether you’re working with an external client or an internal team. This can present challenges for the SEO practitioner:

  • How do you justify the cost of your SEO recommendation to decision-makers?
  • How do you prove an ROI for recommendations that have not yet been implemented?
  • How do you ensure your SEO recommendations get prioritized and implemented?

SEO teams (like the technical SEOs at LOCOMOTIVE Agency) develop many different strategies to address these challenges. Split testing is arguably the most effective. 

With split testing, you can test out your recommendations and get real results. Then, you can use these results to show the expected ROI, justify the costs of implementing your suggestions, and convince decision-makers to prioritize them.

What Is Split Testing?

Split testing is a controlled experiment that enables you to test changes to your website on a smaller scale before implementing those changes more broadly. This is also referred to as A/B testing. 

To perform a split test, you implement your changes on a set of pages called the test group. Then, you compare the results to an unchanged set of pages called the control group. You can use what you learn to guide your strategy moving forward.

SEO A/B testing used to be much harder to do before the advent of split testing tools like SplitSignal. Now, however, it’s as simple as coming up with a hypothesis, setting up a line of JavaScript, and executing your split test using the tool.

If you’ve never performed an SEO split test before, you can find out more in this Split Testing 101 article.

Why SEO Split Testing?

In short, split testing helps you gather concrete evidence to back up your SEO recommendations. When you want to make changes to your client’s site (or even your own), the evidence you gather during a split test can help you demonstrate the value of those changes and convince decision-makers to implement them more quickly. 

This is true whether your test results are positive or negative:

  • Positive results: Your proposed changes may be prioritized. For example, if your test results in a 50% increase in traffic, your suggestions are likely to be a pretty high priority.
  • Negative results: You save yourself and your client precious time, effort, and budget. You also avoid implementing something that isn’t going to work. 

Split testing can also help you make more effective SEO recommendations. After all, what works in one vertical may not always work in another. Testing can help you check your work so that you’re always making suggestions backed by evidence.

6 Best Practices for SplitTesting Your SEO Recommendations

Whether it’s for an agency client or your site, each split test you perform should be unique and tailored toward the specific goals you’re trying to achieve.

Below, we’ll explore six best practices to help you run more impactful A/B tests.

1. Define Your Split Test With a Hypothesis

Before you perform a split test, you need to know what you’re testing. 

In other words, what impact do you think your proposed changes are likely to have on your website? The expected result of your changes is your hypothesis. Your split test checks to see if your hypothesis was right.

For example, let’s say you’re working with an eCommerce client who wants to improve their performance during the 2021 holiday season. Based on previous experience, you believe that adding “Holidays 2021” to your client’s meta titles would improve the click-through rate of their products

With a split test, you can test your new meta titles on a smaller scale before you spend lots of time making the change across the rest of the site.

Be careful: if your test is too narrowly defined, you may not get enough data to draw any helpful conclusions. Instead, to make the most out of your split tests, try to focus on high-impact elements within your website (e.g., title tags, heading elements, meta descriptions, etc.). 

Once you’ve clearly defined your hypothesis, you can set up the rest of your experiment.

2. Make Sure Your Sample Size Is Large Enough

Like any statistical test, split tests need a sample size large enough to be able to draw valid conclusions. The logic is similar to why you’d rely on a 5-star rating from thousands of restaurant patrons versus one 5-star rating from a single visitor. 

Since this is an SEO test, your sample is most likely going to consist of traffic or clicks to your pages from search. For an effective test, SplitSignal recommends using at least 300 pages per test group, which have earned at least 200 thousand clicks over the last 100 days.

Here are a few things to keep in mind when choosing test and control groups:

First, your group of pages needs to generate enough traffic that you can get a large enough sample.

Second, both your test and control group need to be comparable to one another, meaning they’re made up of similar types of pages and similar intent. This way, any search algorithm changes are likely to impact them similarly, so that you can still draw valid conclusions. 

Third, both groups need to be diversified. This can generally help you avoid unexpected fluctuations in traffic that could impact your sample size. 

3. Communicate Early with Other Stakeholders

If your SEO test could impact other internal or external stakeholders, ensure they know it well in advance. This could have two major benefits: 

  • Fostering cooperation between teams or across agencies
  • Avoiding unnecessarily invalidating your test results

For example, let’s say you’re going to run an SEO test where you alter the heading element on a series of product pages. When you reach out to the CRO team, you find out that, halfway through your test, they had also planned to update some of those headers differently. 

Because you told them about your test, they decide to delay their page updates until after you get your test results, in case your findings impact their strategy. If you hadn’t reached out to them, they might have gone ahead with their original plans and rendered your split test invalid. 

Always maintain an effective line of communication with all involved parties. When in doubt, it’s better to over-communicate than conflict later and create more work for everybody.

4. Know Your Timeline and Plan Accordingly

It will take time and volume to reach a statistically significant result. According to SplitSignal, 21 days is sufficient for most tests. That said, your specific needs might vary. The higher the volume you want to target, the more time you’ll need to budget for it.

Before you run your test, take the time to review how much traffic your test group typically earns and use that information to estimate how long it will take you to achieve your target volume. 

Then, as you communicate your plans with other teams and stakeholders, ensure they’re aware of your expected timeline. This way, they can coordinate their activities and maintain reasonable expectations. 

5. Focus on High Impact Pages

If you want your test results to make an impact: 

  • Focus on pages that have a high impact on your client or your organization
  • Align your tests with the goals of the content being tested 

Money pages will probably get the highest priority from an implementation standpoint, so it may be useful to focus your efforts here. Money pages could include eCommerce product and taxonomy pages, pages related to lead generation, or other content directly related to sales.

Pages like this often use templates, which can be an advantage when it comes to performing split tests. Templatized pages typically cover the same types of content in the same way, making it easier to compare them.

By focusing on high-impact pages like these, you can more easily demonstrate the return on investment for your client. Moreover, you can get your recommended changes prioritized by showing results that align with your client’s goals.

6. Use the Right Metrics for Your Test

Finally, make sure you focus on the right metrics when you evaluate your split test results. For SEO, these would be traffic metrics like clicks, click-through rates, and impressions. Other, more conversion- and engagement-focused metrics should be avoided. 

This may seem controversial initially because qualified traffic and conversion are so important, often integral to a company’s goals. However, this is outside the scope of an SEO test. Split tests for SEO focus primarily on how organic search algorithms react to site changes, NOT to user behavior after they arrive on your site

If your pages are set up correctly and include the right content, Google typically knows the types of traffic to send to your pages organically for a conversion. If your pages aren’t written well for conversion, then you should address that issue before you start an SEO split test.

Launch Your Split Test Campaign

To run a successful SEO split test campaign, you must first understand the best practices to follow when designing one. Then, establish your hypothesis, identify the best pages and metrics to use to test it, and set up your test. 

A/B testing can be complicated. Testing tools like SplitSignal can automate some of this process, helping to guide you through the process of creating a more effective test. Then, you can use what you learn to elevate your SEO strategy for yourself or your clients.

Learn more about running an effective split test, then start running yours today with SplitSignal.

Share
Author Photo
Director of Agency Channel Sales for Semrush‘s Enterprise Product Division
More on this