SEO A/B testing, sometimes called SEO split testing, tests changes to your website and measures the impact of those changes on organic traffic via search engines. By comparing the performance of variants (changed pages) to a control group (unchanged pages), you can account for meaningful improvements with real data—and not just guess at what might improve your pages’ and site’s performance.
In this guide to SEO A/B testing, we’ll break down the basics of split testing and show you real-world results. We aim to inspire you to run your own tests—and reach out for a little help from RooLabs if you need it.
How Split Testing Can Help Your Site
Optimized websites and pages are now commonplace on the World Wide Web. In the previous internet era, you could often stand out among your competitors with a few simple tweaks. Today, everyone from the plumber across town to the e-commerce behemoth—and everyone in between!—has a website “optimized for SEO.”
With such high stakes on the SERPs, you need something to help you stand out in overcrowded spaces. SEO A/B testing can help you do just that.
Improved Rankings
SEO is all (or at least mostly) about improved rankings—and so is SEO A/B testing. The difference is that split testing takes us from hunches and best practices to hypotheses and conclusions that provide us with data-backed conclusions on what works best in the real world.
Once you have hard data on what works well for your website, or at least the group of pages you are testing, you can apply data-driven updates to a broader section of your website with confidence that what you are doing isn’t just a shot in the dark but a solid path to success.
Enhanced User Experience
If you’ve been tracking Google algorithm updates (and leaks!) for any length of time, you realize that the search engine is increasingly user-centric. Even if they don’t directly reveal the impact of user signals on search rankings, we know, generally speaking, that what benefits users also benefits our positions in the SERPs.
So in theory, tests that improve your performance in Google will also benefit your users and vice versa. What’s good for the user is good for Google too!
Competitive Advantage
All your competitors are “doing SEO.” How many of them are running SEO A/B testing? With “built-in” SEO on pretty much every web platform (and “SEO optimization” going to the lowest bidder on freelance sites), you won’t stand out any longer with run-of-the-mill tactics. SEO A/B testing can offer you a competitive, data-driven advantage in a marketplace where SEO is now the cost of admission rather than a luxury.
SEO A/B Testing vs. CRO
Generally, SEO testing focuses on getting users to pages, while CRO is about converting those users once they’re there. SEO A/B testing targets what must happen to get users to a page; CRO addresses what happens after they land on the page.
Here’s the bottom line: CRO tests split the audience into control and variant groups to measure conversions. SEO tests split pages to evaluate performance for all users, including search engines like Google.
The crucial piece to remember for SEO testing is that we can’t serve Googlebot variations of the same page. Having more than one page dedicated to the same content or purpose would cause duplicate content and indexation issues—and, ultimately, defeat the whole point of SEO testing.
Key Elements of SEO Split Testing
SEO A/B testing is more than just applying hunches and best practices and hoping for a line that moves up and to the right in Google Search Console. Split testing is a structured process focusing on specific elements of a random control group of (often) similar pages.
Control and Variant Pages
You can’t run an experiment without something to test against. In SEO A/B testing, you’ll work with two groups of pages: control and variant. The control group remains unchanged and serves as the baseline. The variant group has the changes applied. With something to measure against (the control group), you can isolate the effects of your changes and objectively demonstrate the performance changes.
Grouping
Try to balance your control and test groups evenly to get meaningful results from your split tests. For example, once you’ve isolated the types of pages on your site you are testing, say location pages or product category pages, you might be prone to run tests on underperforming pages. A variant group of underperforming versus a control group of pages that are already performing as expected before the test would skew your results and obscure the actual impact of your changes.
Elements to Test
Essentially, you can test anything that impacts SEO—which, nowadays, we know is pretty much everything. Here are some more common testing elements:
- Title Tags: Testing changes in title tags, like replacing your brand name with a CTA like Call Today! or adding keyword variations, can reveal what drives more impressions and clicks.
- Meta Descriptions: Meta descriptions can have a measurable impact on CTR. Testing full rewrites or even rewrites of one aspect of a meta description (including an offer, using title case, utilizing all caps) can provide you with meaningful insights.
- H1s: What happens when your H1s are an “exact match” for the keyword you are targeting? What about when they include more than just the target topic? Measure and test to determine what your H1s ought to be.
- Internal Linking: Internal linking is a bit more challenging, but set up parameters for internal link placement, frequency, and anchor text to see what happens with the variant pages.
- On-Page Content: Adjusting content placement, length, or keyword usage can provide insights into what ranks best. Just don’t forget to make measurable, repeatable changes—not an entire rewrite that’s difficult to measure.
How to Run an SEO A/B Test
With all that context in mind, let’s review how you might approach an SEO split test.
1. Choose elements to test.
Ensure you isolate an element to test. You’ll never know what change made the biggest impact if you test too much at once. So instead of testing general page updates, choose specific elements of your pages:
- Title tags
- Meta descriptions
- H1s
- Other headings
- Internal linking
- Content placement
- Content length
- Keyword usage and distribution
Don’t forget that you can always test more in time. Overtesting will dilute your results and minimize the usefulness of any data you collect.
2. Formulate a hypothesis.
A strong hypothesis based on your SEO objectives will set the right direction for your testing. Be specific and measurable.
You might want to try Conversion.com’s hypothesis framework, which involves what you know, what you believe, and how you’ll test. For example:
- We know that Google uses H1s (or at least bigger, bolder text on a page) to determine keyword and topical relevance.
- We believe that updating our H1s to more closely match the page topic will demonstrate higher page relevancy to our target keywords.
- We’ll know by testing pages with refined, closer-match H1s against a control group of existing H1s and observing and measuring organic traffic to the variant and control groups, respectively.
3. Divide pages into control and variant groups.
Ensure you choose a group of pages that are as similar as possible. You don’t want to test H1 changes for blog post titles against product category pages or local landing pages.
Once you have a good test group, you can divide it into a control group and a variant group. The control group will remain unchanged during the test, while the variant group will receive the changes stated in your hypothesis. Randomly assigning pages to these groups is crucial to avoid skewing your test results.
4. Implement changes and wait.
Apply the changes specified in your hypothesis to the variant group—and only those changes!—while keeping the control group unchanged.
You may be tempted to make other changes while updating your variant group or examining your control group. Don’t do it! Remember that you are attempting to isolate and measure the impact of specific changes, not just “page updates” in general.
After that, you… wait.
5. Measure and analyze results.
After you’ve deployed your test changes, monitor the performance of both the control and variant groups over a set period, perhaps 2–4 weeks. Depending on what you changed, you might see changes sooner, but be sure to stick to the duration you set at the beginning of your test. Key metrics include:
- Organic Traffic: The number of clicks and impressions arriving via search engines is perhaps the single most important metric to measure.
- Rankings: Fluctuations in positions for target keywords can help you assess impact, especially if CTR is lower across a crowded SERP.
- Click-through Rate (CTR): Measuring CTR can help you normalize your findings because CTR isn’t impacted by search volume fluctuations like raw organic traffic.
Observe and analyze your metrics, and decipher whether your hypothesis holds up in the real world.
Challenges to Split Testing
SEO split testing can prove challenging to measure. SEO has so many volatile variables that accounting for your specific changes may ultimately be more difficult than not.
Insufficient Traffic
The “experts” will tell you that your test will not achieve statistical significance unless there are tens of thousands of visits to your test group pages.
But we say: Who cares about statistical significance when the traffic numbers are up because you discovered measurable changes that improve your pages?
Don’t mishear us! Statistical significance shows that relationships between the variables being tested aren’t random. In other words, if your tests are statistically significant, you can be confident that your improvements are caused by your SEO changes, not random chance. Think of it like flipping a coin: flipping it five hundred times instead of five increases the chances that your results aren’t random but reflect real-world probabilities.
Our emphasis here is on practical significance, or what makes a difference for businesses, regardless of whether they have the traffic to numbers to attain statistical significance. We’re glad of statistical significance; we just won’t want smaller websites to dismiss testing because they don’t have the traffic to reach statistical significance.
Too Few Pages
Like traffic, the bigger the group of pages, the better. You should still run experiments, even if you are only testing across a few pages. But again, with a smaller sample size, you’ll have to hold loosely the conclusions you draw.
Unpredictability of SEO
SEO is not a static enterprise. With algorithm updates and tweaks happening all the time, it can prove challenging to know whether you can pinpoint your changes as having a causal effect on the increased (or decreased!) traffic you are experiencing. You may also be in a situation where sitewide changes are beyond your control, even though, theoretically, those changes may impact your test.
The volatility of SEO shouldn’t stop you from testing. You may need to hold your conclusions loosely in the end, but if your changes imply a measurable positive impact, we say test and update away!
Tools and Resources for SEO A/B Testing
Once you start diving into SEO testing, you’ll quickly realize you need tools to help you run your tests.
We highly recommend SEOTesting.com for your tests. SEOTesting has built-in tools that enable you to conduct split tests right out of the gate.
If you have higher traffic numbers and a larger website, Search Pilot may be the solution for you. It offers an enterprise-friendly platform to conduct A/B tests that will move the needle for your business.
Test Away to Improve Your SEO
By using SEO A/B testing to make data-driven decisions, you’ll improve your rankings, attract more organic traffic, and gain an edge over your competitors, who, at this point, are all “doing SEO.”
Start small. Test frequently. Don’t be afraid of where the data leads. You may find that what works for you creates friction with conventional SEO wisdom. Oh well. If you have the data on what works best for your site, that’s what matters.
Test away, home in on what works, and watch the benefits pour in.
Did you know that RicketyRoo has an SEO testing division called RooLabs? We are constantly testing across a breadth of industries to establish better SEO practices. Reach out to us if you’re interested in help with local testing.
