by Brian Harnish

How to Improve Your Website Conversions Using Split Testing

In the past several years, SEO has become a truly integrated marketing machine. This means that many different facets of marketing are now essential for improving SEO. SEO is not its own little technical silo anymore. In order to gain the most improvement in SEO, you have to always test, test, and test. And test some more to confirm that the change will be effective. And I am not talking about Google’s algorithm, although building websites to test Google’s new algorithm changes is always a good idea. No, what I am talking about is testing different alterations of the website on your clients so that you have a solid idea of what certain changes will do for your conversions. To do this, you have to perform testing of different layouts, site design, and potential SEO changes. If you don’t, then you really will have no idea if that design had a positive improvement or not.

A/B vs. Multivariate Testing and Why They Are Important

This kind of testing involves testing different layouts against the original before making the change to the new one to see if it yields any improvement at all. This kind of testing is called A/B testing. If you are testing changes on multiple elements of the exact same design (or multiple designs), then you are engaging in multivariate testing. Say Company A decides to change their website, but they have yet to engage in any kind of testing whatsoever. They have just spent $5,000 on a design that someone says has yielded significant conversions. However, these conversions were never nearly what the design before it had. In fact, this new design has resulted in a loss of 5 conversions a month. If the company had invested in testing the design before implementing it, they probably would be out $5,000 still for owning the design, but at least they didn’t lose 5 conversions a month that cost them a lot in profit and conversions. However, Company B engages in testing regularly before they completely change their website. They come across a design that looks pretty and beautiful, with a lot of widgets and additional functions their current design lacks. True to their nature, they engage in the testing of this design without a second thought. Testing reveals that switching to this design caused a conversion drop of 15 conversions a month, resulting in 15 lost sales per month. They decide to stick with the current design until a suitable one with better conversion improvement comes along. Thanks to testing, they didn’t just decide to put up a design that resulted in significantly lost profits to the company along the way.

How to Perform A/B & Multivariate SEO Testing

There are a few simple steps to performing A/B and Multivariate Testing:

Decide What You Want To Test

Do you want to test an entirely brand-new layout? Do you want to test the change of the color of a single button? Discuss with your team what you want to test and how you want to proceed with this testing.

Decide How You Want To Test

Do you want to use A/B testing? Or Multivariate? If you want to use A/B testing, it’s a good idea to plan how long you plan on running the test for, how long it will take for proper conversions to start to take place in your industry niche, and for how much of a segment of your traffic you want to test for. It’s typically a good idea to have at least 10,000 unique visitors before you start considering testing in order to eliminate confusing variables, to minimize errors, and inconclusive results.

Decide How Long You Want To Test

The length of testing time will vary between industry and how conversions take place on your website. For new designs, you will want to test them for longer than a month in order to gauge how this design performs against prior year benchmarks. If you test this design for a month, you won’t have a proper length of time in place to gauge how effective it is. For example, if you had 45 conversions in February of last year, but this new design only gets 25 conversions this February, then you can get misleading results. It’s a good idea to run such a test for around 3 months minimum to gauge its maximum effectiveness against prior similar months in your website’s history. Of course, don’t forget to take into account slow periods in industry shopping times, and busy periods. Depending on your industry, this can fluctuate wildly and such minimal lengths of time can lead to misleading testing results. So, to recap:
  • Test against prior years.
  • Test new designs for longer than 30 days.
  • Take into account whether or not you are in a slow or fast period in your industry.
  • Decide what you want to test.
  • Decide how you want to test.
  • Decide how long you want to test.

Identifying Metrics that Management Cares About

SEO has many metrics. From conversions to bounce rate, to visits, to unique visitors, to pageviews, average visit duration. For those in management who are not SEO wizards, then looking at these metrics can be confusing. However, it is important that you can understand some of them so that you know how your SEO efforts are really affecting the website and your company’s bottom line. First off, let’s go through some of the metrics that are important to determining how the website is performing in the SERPs: Impressions – This is how many times a website has actually been viewed in the SERPs. Clicks – This metric identifies how many times the site has been clicked on in the SERPs. Neither impressions or clicks are typically satisfying because they deliver misleading results because of their arbitrary nature. In other words, if you have 300 impressions but 150 clicks, clicks be misleading because clicks do not account for the single person that’s clicking on the site. If the same person went to your site via the SERPs came back, went to another site, and came back, those metrics will report those two clicks. They don’t differentiate between IP addresses or whether or not it’s a single person clicking on that SERP link. Visits – This is another arbitrary metric that isn’t quite accurate. You can have single visitors being reported as multiple visits by Google Analytics. This is because GA does not report visits as single visitors. In fact, these are generally referred to as sessions because of the fact that GA uses 30 minute cookies to track the users’ visits in 30 minute “sessions”. This can also be misleading in terms of reporting proper real visitors to the site. Unique Visitors – The metric management will want to care about most is the Unique Visitors metric. This will help management understand better how much a specific SEO effort is contributing to the site from a visitor standpoint, and is one of the most useful metrics you can include in your reporting. Bounce rate – The bounce rate of pages is another useful metric. With the aid of interpretation from the SEO, this can tell management how pages within the SERPs are performing. Are they performing well? Or are they not performing as well as they should? However, it should be noted that there are a few exceptions to this rule. If a page has an extremely high bounce rate, it will be necessary to check the logs to ensure that they are not being visited by a ton of bots rather than people. Bots can muddy up the bounce rate here quite a bit so it’s important to ensure that this metric does not take into account bots but is actually reporting real people…at least, as real as possible. Time on page – One more metric that is useful to include in reporting to management is how much time people spend on the specific page. This metric, tracked over time, will be able to help you decide which pages are complete junk so you can trash them, and which pages to keep to improve them.

A Brief Introduction to Analytics & Analytics Metrics

Google Analytics is a powerful reporting platform that can be used in your daily, monthly, and yearly website reporting. It’s one of the most often used analytics platforms and provides just about everything you would want to know about your website visitors. There are a wide variety of metrics that GA reports on. Including: Demographics – the language and location of your users. Behavior – New vs. returning visitors, the frequency and recency of visits, and user engagement. Technology – This metric allows you to see the types of browsers and operating systems that are most frequently used to access your site, and the network that is used to access your site most often. While most of these can be found in your server log files, GA has a nice, pretty, organized layout that can quickly be exported to Excel, which is not always true of every server log file. These metrics can be useful in determining which people to tailor your website to, and how to plan changes and maintenance across a website to make sure that these people are your happiest customers. You don’t want your potential clients to run into issues with your website on their machines! Having many hands in a website can ruin the pages from time to time, so having a maintenance guru on staff can be a boon for your marketing department. Mobile – Using metrics in this section of GA can help you determine whether or not someone used their desktop or a mobile device to view your site. And it goes into such detail as to which mobile devices accessed your site. This metric is important to figuring out whether or not you should expand into mobile responsive design yet. However, a responsive design is always a good idea because such a design tends to be SEO friendly and will enable your site to be viewed across a wide variety of browsers and platforms. Custom Variables – You can set custom variables in GA to find out what your metrics are in a wide variety of different scenarios that may not be provided by the GA default. Visitor Flow – This section of GA can tell you, in detail using a nice flowchart, where visitors on your site came from. Whether they are from the United States, China, UK, or other countries, the starting pages they landed on, and the first interaction they performed.

Google Analytics – Traffic Sources

Google analytics provides some great metrics to use to report on your SEO traffic. All Traffic – This metric provides an overview of all the traffic to your site and does not discriminate between traffic types. Direct Traffic – This metric provides information on which of your URLs people type into the browser. So, the URLs people remember the most, what they bookmark, and such URLs will count in your direct traffic. Referrals – Referrals is all traffic that is “referred” from other websites to your website, and does not take into account metrics provided under direct traffic. Search – Organic & Paid Search traffic Organic – You will be able to see who found your site through the free Google organic results through this metric. It’s very useful for seeing your site’s general performance in the SERPs, although it has become muddied as of late by Google’s egregious use of “not provided”. If you can successfully work your way around this not provided hiccup, you will be able to see most of your organic traffic. Paid – This option allows you to see all of your traffic that comes through via the paid search campaigns you are running.

Search Engine Optimization

Now we get to some of the more interesting SEO analytics sections! Queries – This allows you to see which search keywords people type in that actually bring them to your site. You will need to enable Google Analytics to access your Google Webmaster tools account in order to see this information. Landing Pages – This option will allow you to see how your website pages are actually performing. This can help your company answer a lot of questions about what to do about pages that are performing or pages that are under performing. In this section you can find information about pages that provide an introduction to your site, the pages that different traffic segments land on, blog posts that drive the most traffic, and see the overall performance of your landing pages.

Google Analytics – Content

Using the content section in Google Analytics provides a significant amount of information on many of the different pages of your site and how your visitors interact with them. Based on the data, you will be able to better figure out how well the content addresses your users’ queries and intentions. Each of the page sections in Google Analytics content > site content tab provides information on all of the pages of your site such as: the pageviews occurred, actual unique pageviews, the average time on page, the bounce rate of the page, and the percentage of people who exited the site from that page. If you are performing a website audit, using the data in this section will help you determine which pages are actually useful and performing from the point of view of the user. Site speed – The information in this section is useful for determining the pages that are underperforming. For example, the data displayed in this section’s overview will show you average page load time, average redirect time, average domain lookup time, average server connection time, average server response time, and average page download time. Going into further sections under site speed will help you diagnose issues with specific pages that you may want to improve. It will help you find one of those pages that have an image that’s taking too long to load, or other issues that are occurring on the site. Going through this section on a large site with a fine-toothed comb will help you diagnose issues that you might not otherwise discover.

Content Experiments

Here, we get to the section of Google Analytics that allows you to perform A/B and Multivariate Testing!! In Content experiments, you can set up these kinds of tests in order to determine what kind of changes to make to a specific page that will be effective. You will also be able to test an entirely new design by using the features of GA’s content experiments. It’s up to you how much and how you want to test using this section. It will help you test landing pages, your final goal conversion page, as well as the pages that your website visitors will encounter along the goal funnel. By learning how to use the basic tools at your disposal, and how to correctly integrate them into your split testing methodology, it will be easier to create accurate reporting for your clients.