The Only World-Standard SEO Software

Download Now
SEO PowerSuite
SEO PowerSuite Hot-new version
Supported OS

Best landing page split testing tips from the PROs

February 18th, 2013 | | Posted in category Internet Marketing On Page SEO Website Usability

If you are reading this article, you probably know what a landing page is. At least we hope so, because we can't wait to share our tips with you instead of telling you boring basic things which are two clicks away from you in Google.

Well, while most marketers know what a landing page is, not all of them clearly understand what split testing is and how it can be useful for a client. I believe most of SEO folks are aware of basic landing page rules and build their pages knowing what they do. But just using these tips is not enough if you are not satisfied with conversion or want to squeeze more results from your landing page.

SEO cannot be over, it can be paused or stopped, and there is also something you can do to improve the results. Landing page split testing allows you to understand what works best.

Why use split testing?

  • general recommendations are good, but still they are general. Things are constantly changing, and you have to track them in order to keep up to the progress;
  • your audience is specific. Various techniques work for various audience structure;
  • sometimes you may find that even slight changes bring results in conversion. Depending on your business even 0,1% may appear to be hundreds of dollars.

How to do split testing

  1. create 2 (A/B testing) and more (multi-testing) pages or page combinations and show them to segmented audience;
  2. use any tool to control the experiment and watch the metrics needed to analyze the result. The simplest free option is to use Content Experiments in Google Analytics;
  3. choose the winner page, or continue the experiment if the results were not sufficient;
  4. continue to carry experiments to adjust your page even more;
  5. repeat.

Well, you can find more tutorials on that simple algorithm, and we won't dig into things that were published before by thousands of blogs. I am here to highlight the details that came from my experience and from my colleagues' vast practical knowledge on controlled experiments in e-commerce websites.

Pro tips from Link-Assistant.com Marketers

  • If it is possible, test pages for various audience. This method shouldn't be used if you have a powerful flow of diverse traffic, as you may spend too much effort while segmenting it. But it is good to test landing pages for ads separate of landing pages for organic traffic.
  • Do not use redirecting for showing different pages for the users. If you break the rule, be ready to be penalized by Google very soon. Content Experiments service in Google Analytics uses JavaScript and cookies to perform the testing.
  • Look at the competitors. If you know how split testing is done, you can clearly find out if it is done on a website. It can be identified by JavaScript codes presence on the page and showing different pages to the user. Enter the website several times from ads and SERP, and if they perform experiments, you will notice several tested landing pages at once. Track what they are doing: for how long the test has been carried on, which page eventually won and was left on the website, do they change it from time to time and so on.
  • Read cases on split testing and learn from them. Although many cases are published for commercial reasons, you can still pick some tricks from them. Here is an example of a case studies bunch. Some of them showed that very long pages really work for some audience.
  • The experiment period may vary depending on your traffic. If your traffic is thousands of people, it is OK to test pages for a short time. But, say, you have a hundred of visitors a day, than it is better to carry the experiment for 1-3 months. The less traffic you have, the longer the test should be. Here is a nice calculator that shows the needed duration of an experiment depending on the metrics you have now.
  • Do not make quick decisions, take your time to interpret the results of the experiment. Do not hurry to declare a winner if you just see a page with seemingly high conversion. Try not to forget about all the factors.
  • Minimize actions that may influence the result. Do not do any bulk mailing, contests or actions that may influence the traffic to the pages tested. Try to avoid experiments during holidays or seasonal occasions that may be connected to your business.
  • Landing pages consist of many elements. All in all you have to decide on:
    • call to actions;
    • headline;
    • main text or product description;
    • form’s length and types of fields;
    • layout;
    • pricing and offers;
    • images;
    • amount of the text and length of the page;
    • structure of the page and elements position.

It is wise to consider the winning design first, then choose the text, and after that it is ok to try different variations of separate elements color, position etc.

  • Always consider the country of your potential visitors. The Chinese love the text more. Typical and familiar landing pages for US visitors will not be alike to those for European potential customers. Search for design and structure trends for a particular country.
  • Landing pages for organic traffic and for ads and links may be and are very likely to be different. You may find that your landing page that gets organic traffic from SERPs is showing very low conversion for ads. So study carefully and don't be lazy to create separate landings for various traffic sources.
  • Sometimes your experiments may lead you to change your landings significantly, and they will not even resemble your website's layout. You shouldn't be afraid of that. The main aim of the landing page is to convert traffic, so use any methods for that.
  • If you created your landing pages considering all the guidelines, you are not likely to expect massive effect from your experiment each time you run a test, because most things have been already done in a good way. But that does not mean you shouldn't proceed. Increasing conversion percentage brings you money. And, what is more, the guidelines get old and ineffective, so you have to check their relevance and effectiveness from time to time.
  • Do not fully trust the experts. It is not about no value in their recommendations. Just remember that not all of their pieces of advice are true when speaking about the users. It is better not to predict users’ behavior, but to track how they really react on the webpage. It is also true about our intuition, because it must be checked before applying.
  • Build funnels and check not only separate pages, but also bundles of them. For example, after your landing page the user gets to be registered and visits the page where he inserts his personal data. So check various combinations of landings and registration pages, for example. You may find unexpected results here.
  • Check all the metrics that are available. Do not focus on the conversion only, as other parameters will tell you more about your clients. Look at time spent on the page, bounce rate, form errors, and more.
  • Decide whether you would like to convert new visitors or those who are familiar with your website and business. The approach to their conversion is different. Do not change habitual navigation for the users that are here not for the first time. You definitely do not want to surprise regular users.
  • Run contests and actions after the major changes are made. Journalists and the mainstream user mass should see your changes after they have been made.
  • Run A/A tests. It is the same as A/B test, but here A=B. It helps you to identify discrepancies.
  • Remove robots, bots and spiders while testing. They can generate from 5 to 40% of traffic and confuse the results. Sure, it is not about pages for organic traffic from SERPs.
  • Use various tools for experiments so it does not require manual coding. For example, creating pages using CMS is much more efficient than creating them manually.
  • Say, you've tested two pages and got the results:

Page A, 2556 visits, 12 goal successes

Page B, 2300 visits, 11 goal successes

Here we cannot say exactly which page won. Also the success of the experiment is not evident here. In this case we can use a calculator to count the results. For example, this one.

  • If it is possible, show the same pages to repeat users. Usually the testing tools do it with cookies.
  • Also test all the versions simultaneously. Do not wait for the second version to be tested after the first one is done, because you need the same factors that influence the result.

Controlled testing in e-commerce is a field where marketing, usability, and statistics mix in one. So here you may want to dig deeper into these territories, we suggest reading more information on the strategy of controlled experiments, which is actually specified not only for the e-commerce websites, but for all kinds of experiments, in marketing too.



back to SEO blog


  • http://www.persistentengagement.com/ Franklyn G.

    Great post, always informative thanks!
    I only use A-B testing every so often but only for a short time!

    Sophia Kowalski Like or Dislike: Thumb up 0 Thumb down 0

    Hey Franklyn, glad you like it.

    So how long do you usually test the pages?