Stop Guessing, Start Testing: The Split Testing Guide.

In the recent Beginners’ Guide to Conversion Rate Optimisation that I published on the SEO+ blog, I mentioned how A/B split testing can help you significantly improve the conversion rates of your website or online marketing.
 

Although split testing requires you to invest some time, the pay-off should be more effective marketing and increased profits. One stat I noted down recently suggested that split testing can make a landing page up to four times more effective!

What is split testing?

Split testing is sometimes referred to as A/B testing or multivariate testing, although the latter actually relates to split testing pages that have several variations on one design. Essentially, split testing is a way of looking at the different elements on a landing page (or ad or marketing email) and building up data about how these elements can best be used to meet your conversion goals.

Although a web designer should have a good feel for how visitors navigate through a website and respond to different elements on the screen, such as booking forms or call to action buttons, split testing takes away the guess work.

When you run a split test, you send 50% of your traffic to the original page – often referred to as the ‘control’ page – and then 50% of your traffic to another variation of the page. I would always recommend that you only change one element of the page at a time on the non-control version. For example, if you’re not sure about where to place the call to action, you would move this and only this in each variation to see which position gets the highest clickthrough rate.

You might then do a second split test where you change the colour of the call to action button to see which visitors most respond to. Do visitors associate green with ‘go’ or are they more likely to respond to the urgency of red?

Subsequent split tests might look at the wording of the call to action, the design of your sign-up form, images, text or image placement, design features such as arrows that point to the call to action, and much more.

By changing just one element on the page at a time, you can accurately pinpoint which changes on the page had a direct impact on your conversion rates. Over the course of several split tests, you will begin to build a picture of what the ideal web page should look like for your customers to convert.

Once you have built up a picture of how single elements affect your conversion rates, you may then need to split test how these elements work together. For example, although people clearly like a specific colour for your call to action within your original design, do they still respond to the new colour when the button appears in a different position?

Alternatively, you could create an A, B & C version of your landing page, sending a third of your traffic to each one to test three different versions of the same element, e.g. three different colours for your call to action button, three different headlines, or three different images. This can a great time saver.

Baseline results

You should always have some baseline results to compare your split testing to. It’s a good idea to launch the control version of your landing page first and monitor the traffic and visitor behaviour in Google Analytics. Once you have built up an initial picture of how people are interacting with the page, you can begin the split testing process.

Otherwise, you could monitor your control page for a while, and then create an A and B version of the page to compare against this, allowing you to test two new variations instead of one.

You should always all the versions of your landing page at the same time to ensure that there are no variations in the conversion rates caused by timing.

Know your goal for each page

Before you carry out a split test, you should know exactly what you want to test and why. It is helpful to define a clear ‘conversion’ goal for each page.

• Do you want people to sign up to your mailing list by downloading a free ebook?
• Do you want a website visitor to fill in a booking form or buy a specific product?
• Perhaps you want them to pick up the phone to request an appointment?

Look at every element on your web pages through the lens of your clearly defined goal.

• Does the content communicate the benefits of clicking on the call to action?

• Do the images reflect the content?
• Have you got rid of distractions?
• Do all the design elements show the web visitor what they need to do next?

Get rid of distractions

Split testing is incredible valuable when you’re designing a landing page as landing pages are solely focused on conversions and should always have a single goal.

My advice is to get rid of all distractions on your landing pages – this means no navigation, no sidebar, no footer, and no links that might take visitors away from the page. Try to take away anything on the web page that could let the visitor navigate away without completing the action you want them to take.

What can you test with split testing?

You can evaluate nearly every element of a web page with split testing, as well as marketing emails, e-newsletters, ad campaigns and more. Where there is a measurable goal, there should be a way of carrying out a split test and recording the resulting data.

• Look at the visual elements such as the placement of pictures and videos, or the colours you use.
• Split test text elements such as headlines, calls to action, and product descriptions.
• Explore the layout of your page by split testing the placement and size of buttons, menus and forms.

Although you can split test everything on a landing page, this could be time consuming. Therefore, you might want to concentrate on the big hitting elements of the page such as the:

• Headline
• Call to action
• Sales copy
• Product description
• Sign-up form
• Main ‘hero’ image on the page

Split test duration

One challenge is to work out how long you should run each split test and making sure that you build time for this into your campaign, especially if you’re planning to use a landing page for a time-limited offer.

Run a split test for too short a time and you may not get enough data to decipher which is the best version of the page, but run it for too long and your data may get skewed by external factors, such as competitors’ activities or a big industry news story.

We found a helpful free Test Duration Calendar by VWO that can help you work out how long to run each split test based on your number of daily visitors or the percentage by which you want to improve your conversion rates, among other factors.

SEO and split testing

Before you begin any split testing, you should make sure that your control page has been properly search engine optimised. This means using:

• SEO-friendly URLs: Use your focus keyword for that page in the URL, if possible, and separate any words with dashes, e.g. www.example.com/seo-friendly-urls
• Title tag: Try to use your focus keyword as near to the front as possible
• H1 heading tag: Again, try to use your focus keyword in your main heading
• H2 headings: It’s a good idea to use a couple of H2 subheadings to break up the copy. Try to incorporate your focus keyword and/or other words and phrases that mean the same as your focus keyword.
• Bold: Ensure that your focus keyword appears at least once in your body text and try to highlight it in bold.
• Alt tags: Include a couple of images on your landing page that are relevant to the content and include descriptive alt tags. If you can incorporate your focus keyword into the alt tag, that’s great.

During the split testing process, you will have at least two pages (the control ‘A’ page and page ‘B’) with almost identical content on separate URLs. If you use a free split testing tool like Google Website Optimizer, you can test up to ten variations of a page at the same time.

This has the potential to cause problems in terms of Google penalties for duplicate content. The search engines may also become confused about which landing page variation to index. To make sure you don’t dilute your link juice or get hit with a duplicate content penalty, we would recommend that you:

• Add a ‘No index’ tag to landing page B (and subsequent variations)
• Add a canonical tag to tell the search engines which is the main version of the page that they should be indexing
• Don’t add versions B or beyond of the page in your site map or robots.txt file

Once your split testing is complete, you may well have a different version of your landing page to add to your site. Changing the design and the content of your page may initially affect your search engine rankings. However, on the flipside, of the traffic that does come to your site, more visitors should convert. With split testing and conversion rate optimisation, the emphasis is very much on quality over quantity. It’s worth revising the SEO steps we talked about above to reflect the content of the revised page.

Choosing the right software for your split tests

Creating lots of versions of the same landing page and equally funnelling your website traffic to different URLs can be daunting. The good news is that there are various testing software systems out there to help you manage your split testing and understand the data.

If your budget is limited and you want a free tool, then Google Content Experiments is a popular choice that will let you test up to ten variations of a web page at any one time. However, if you’re not a coder, you may need some IT support to make full use of this tool.

Unbounce is used by some of the world’s most trusted brands and agencies for split testing. The software lets you build stunning, highly converting landing pages without the need of a technical team. Drag and drop any element for fast page design, customise for mobiles, and tap into a range of advanced design features. Unbounce currently offers a 30-day free trial before you would need to move over to a paid plan.

Visual Website Optimizer also offers a 30-day free trial before you would need to sign up to a monthly payment plan for your split testing. This software is easy-to-use and doesn’t require any coding knowledge. You can draw on a library of successful conversion tools highlighted by other split tests, and test any element of a web page. You can even personalise the content website visitors see based on their location or the device on which they’re viewing your landing page.

Optimizely offers a range of free and paid-for plans to enable you to carry out split testing or personalise your web pages. The software is ideal for A/B testing and multivariate testing. The stats delivered by Optimizely are said to be set at a 95% significance level with a 95% chance that you’ll be able to pinpoint the right variation of the page for your customers.

Split testing services are also available from companies such as Kissmetrics, Webtrends, Maxymiser and Convert.

We should point out that none of these businesses are affiliated with SEO+ in any way.

A guide for your future marketing

One of the greatest benefits of split testing is that once you know what landing page has the best conversion rates or what ad gets the highest number of responses, for example, then you have a template for your future marketing and know where your money is best spent, which can lead to massive savings on your marketing budget, as well as bringing in a higher number of new customers.

Split testing isn’t a one-time only process. Design and copy trends change, so you should consider running split tests on a fairly regular basis to keep your landing pages in tip-top condition with high conversion rates.

Do you use split testing in your business? If so, has it helped you make decisions about your landing pages? Have you thought about split testing but been putting it off? I’d love to hear your thoughts in the Comments below.

If this article has given you food for thought, can I ask you to use the social buttons below to spread the word? It takes seconds. Thank you.