
How to Do A/B Tests and How Can They Work for SEO
The competition for the attention of users and customers is getting fiercer every year. This factor has especially increased because of the global pandemic, which has caused a significant part of the traditional off-line business to move online, and it will continue to develop on the Internet.
Therefore, for SEO, content marketing, and marketing in general, it is critical to understand what exactly attracts customers, keeps their attention, and influences their behavior and choices. For this, various tools are used, including A/B tests and multivariate tests. Let’s see what the definition of A/B testing is, find out what its benefits for SEO are, and learn how to work with these tools.
What Are A/B Tests?
A/B tests, which are also called split tests, are the comparison of two versions of webpages, main and tested one that has been changed. For this, site traffic is divided into two parts, each is directed to its version of the page, and after a particular amount of time, conversion rates will show which version works better. Such testing is most often conducted to check changes in design (color schemes) and call-to-action messages.
What is A/B testing for business, and what role does it play in digital marketing? It is an effective and widely used tool for determining which of the design options or elements of the site, content, and methods of its presentation will be best perceived by the audience, bringing more clicks and actions (subscriptions, reactions, orders, purchases, etc.). The main advantages of the method include:
- The small number of studied variables makes it possible to get accurate data quickly, and the result will be indicative even for sites with few visitors and low traffic.
- The possibility to accurately analyze specific parameters (for example, a new title option or the color of a call-to-action button) and evaluate their effectiveness.
- A credible way to convince your skeptical clients or partners of the need to change is by demonstrating a specific result.
You can evaluate the effectiveness of a method by a series of A/B test examples. For example, in 2018, Ubisoft, the French computer game developer, increased the lead generation rate by 12% using this method. For Ubisoft, the principal indicator for assessing the success of the site and user experience is the conversion and dynamics of lead generation. The Buy Now page for the gaming brand For Honor fell far short of these indicators.
Back then, it looked like this:
After having analyzed user information, click maps, reviewing feedback, and conducting a survey, Ubisoft realized that the problem was that the process of buying the game was too complicated. They decided to redesign the page, reduce the number of steps, and simplify the buying process. The new version looks like this:
VWO
A/B testing was carried out for three months and showed that the conversion of the updated test version significantly exceeded the indicators of the old site and increased from 38% to 50%. The overall increase in lead generation is 12%.
The illustrative case of the use of A/B tests is the company WorkZone that has increased conversion on the client questionnaire page by 34% due to changing its color scheme. The page used to look like this:
VWO
Then, Steven Macdonald from WorkZone decided to replace the section with reviews from famous brands from colored logos to a black and white version. They did this to prevent the banner from drawing the attention of users, for it not to be associated with buttons for a click, etc. The changed variant looked like this:
VWO
The result was impressive: in just 22 days of ab testing, the number of users who filled out the form increased by 34%, which demonstrated the effectiveness of the chosen strategy to change the appearance of the page.
It is possible to conclude that the use of A/B tests is a very productive method for determining the effectiveness of the changes being made. At the same time, having a lot of traffic, results can be obtained very quickly, that is why some large sites use this tool regularly. They run tests sequentially on different parameters without resorting to more complex methods of simultaneous testing for more variations and changes.
What Is a Multivariate Test?
It is an elaborated variant of the tool — a/b multivariate testing. The main principle is similar to the one described above; however, when working with the multivariate test, you can compare more variations and also check how they interact with each other.
Like in the case of A/B, the traffic is distributed among the elements that are compared, but given the increased number of compared parameters, this technique is only suitable for sites with significant traffic. The use of a multivariate test is especially relevant for pages with several dynamic elements:
- Registration table, quiz, etc.
- A heading that has to attract attention, describe the product, brand, etc.
- Text, image, or video content.
- A button to lead, confirm, place an order, etc.
In this case, it is possible to enter these parameters and different options for each, and the tool will distribute traffic, carry out a comparative analysis, and define which options yield the most significant response from users, as well as how they interact with each other. As a result, it is possible to carry out optimization focused on user experience and, as expected, get a positive outcome in the form of increased conversions.
For example, the Dutch branch of Hyundai has achieved impressive results by optimizing a site using a multivariate test.
Initially, their page looked like this:
VWO
Setting the parameters for multivariate test, they have chosen such sections:
- The updating of SEO-friendly text. The essence was to create texts based not only on the principles of SEO but also on the versions that yield a higher conversion.
- To add call-to-action buttons.
- To use one big car photo instead of thumbnails. For the test, six combinations were developed (2 for each section). The result of the research was that the page was optimized in all respects: SEO, buttons, and a large spectacular photo of a car. The final result looked like this:
VWO
The site CTR increased by 208%, while the conversion in the form of requests on brochures and test drives raised by 62%.
Most certainly, multivariate and A/B testing marketing is most spread. But what does A/B testing mean for SEO? As you can see from the example with Hyundai, the results of the multivariate test were also used to optimize SEO texts taking into account the conversion that speaks for the user experience. Therefore, these tests can also be considered SEO tools for obtaining reliable information about what interests and can hook the target audience. Let’s look at how to A/B test and analyze the peculiarities of working with the testing of sites and pages.
How to do A/B Testing: A Guide for Beginners
Before we start, we need to decide which types of content this testing will be most suitable for:
- The headline is the first thing users see when they come across your content. So, it is essential to make it attractive, vibrant, eye-catching, and, at the same time, relevant, and SEO-friendly. Therefore, using A/B tests, you can see which headlines most interest your audience.
- Text content: When creating text content, it is critical not only to use keywords and adhere to the relevance principle but also to find a tone of voice that will appeal to your audience. By analyzing the test results, you can better understand the audience and how you need to convey your messages to them.
- Call-to-action: It is crucial to select the right words to get people to do what you expect. A/B tests are among the most effective tools for creating CTAs.
- Images and video: Today, the content that is of the most interest to users contains images and videos. When choosing a visual range for placement on a page, it is worth testing the variants you have selected and proposed to get the maximum response.
- Content volume and depth: It is helpful to know how deeply the target audience is ready to dive into the topic. You can create two options for the content, i.e., short and long versions, and see which version is better perceived by the audience.
- Product pages and description of products. Style, amount of information, manner of presentation, design, etc.
Step-by-step A/B Testing Tutorial
- Conduct thorough research of your resource to find out everything about traffic, the number of visitors to your site or page, the time they spend there, the rate of conversion and its dynamics, etc. Based on this, we move on to the next step.
- Elaborate a hypothesis that, as you think, will lead to an increase in user interest and, therefore, conversion. It is best to make assumptions based on data and research that is relevant to your niche. For example, knowing that users place a lot of emphasis on the visual content, it is possible to assume that the improvement of the design will improve the user experience as well. Also, you can survey the target audience to understand what exactly you have to fix.
- Define the goal: what you want to analyze will be the starting point for choosing the parameters and elements to be tested.
- Select a control page to test and compare with variation page metrics. For example, you want to decide on the format for presenting text content: in a strict business style or an easily readable manner. In this case, what you find preferable should be the original page, and the variation page should be an alternative option.
- Create variations, such as the alternative versions of pages that you will test to confirm or disprove your hypothesis, and also test the control page. For example, if you want to understand why people do not answer the questionnaire on the site, try to consider several versions with the help of variations (for example, it has too personal questions or there are many fields, etc.)
- Choose only one variant from the elements you will be testing for each variation. Naturally, when working with a page or content that you want to optimize, you will need to check several different elements and parameters; however, to get the most accurate results, you need to conduct separate testing with its test page for each of them. For example, if you want to optimize the CTA button, both the content and the look and color of the button, then you should check these parameters separately. Change the appearance of the button and start testing. After you have made sure that most people choose the red button over the green one, proceed to the text testing, etc.
- Choose suitable A/B testing tools. For this analysis, there are paid, and free services, such as Google A/B testing. Before working with this tool and conducting a Google split test, it is necessary to set your Google Optimize account:
Then, accept the service terms of use and go to your container:
After, type the name of the test (experiment) and the URL of the site:
Then, proceed to the start of the process, having already added the page-variant:
Also, through Google Optimizer, it is possible to conduct Google multivariate testing. For this, being on the Experiment page, click the + button and select a Multivariate test.
- Be patient and wait for long enough for testing. How long should the test take? It is perhaps one of the most frequently asked questions, and there is no single answer to it. It all depends on your audience, niche, orientation, and traffic. In the practical examples given in the article, in one case, the testing lasted three months, while in the second, the success of the hypothesis was confirmed within 22 two days. In any case, this is not a quick process, and it should go in parallel with other activities for the development and promotion of the resource.
- Check not only significant but also possible minor changes. It can be done during a break between solving global problems because sometimes, even the smallest details can be decisive for increasing the conversion. Background tone, title size or font, etc. also affect user experience and perception.
- Analyze the results you have received. The effectiveness of the testing is largely contingent on this point: it is essential to take into account all the obtained indicators, calculate the difference and the percentage ratio of user reactions on control and variation pages (for example, if it is a difference of 5%, one can conclude that the changes play no role). If your test was successful, for example, the hypothesis was confirmed and brought you a significant increase in CTR and conversion, it is worth analyzing why people behaved this way, etc. If the testing did not yield the expected result or even turned out to be unsuccessful, it is even more critical to understand what went wrong and how you need to move on.
Another interesting aspect is machine learning ab testing, which is often used by marketers of large companies. The goal is to test the results of machine learning outcomes through a user behavior study. Often, these studies help identify errors in the work of artificial intelligence and adjust the strategy.
A/B Tests and SEO: What Are the Possible Pitfalls and How to Avoid Them
Even though A/B tests are valuable for SEO, it is crucial to remember that improper testing and making mistakes in the process can harm a site in the eyes of search engines. Let’s review how testing can hurt SEO and how to avoid this.
The main problems that you could experience in the process of testing:
- Page duplication: The presence of several nearly identical pages that are available to search engines can create duplicates if this is not regulated and managed.
- Cloaking: If the test pages are very different and during crawling, the search robots do not see what the users see, this can be perceived as a prohibited method of cloaking and lead to the loss of positions in the SERP and Google sanctions.
- Incorrect redirects: If different URLs are used for research, the main and variation pages could be confused.
It is necessary to be attentive and follow particular rules to ensure that the above does not happen:
- Do not create test pages that are too different from each other. It is more applicable for multivariate tests when many elements are checked at the same time, and the segments of the site can vary greatly. It is important to keep track of the relevance of all pages as small differences during the test are typically perceived by search engines.
- Use the tag rel=“canonical” if you work with different URLs during testing to differentiate the main URL. This tag will not serve as a guideline but will hint to the search engine the location of the main page. Read more in the article Rel=Canonical Tag.
- Make sure that Google sees and works only on the page you need to rank. Double-check that it falls into the field of view of the search engine. If, for example, you deleted the part of the key content in the variation of the main page, your positions in the search results may be influenced.
In conclusion, it is worth mentioning that the use of A/B tests and multivariate tests is certainly a useful practice due to which you can evaluate user signals, identify shortcomings, or, on the contrary, strengths in strategy and content, as well as to conduct effective optimization. But when applying them to solve SEO problems, it is crucial to remember safety precautions if you want it to become another tool in your arsenal.