31 Jan 2012

How to set up a successful A/B test


With many SMEs and brands focusing on SEO and social media optimisation these days, actual website usability and conversion optimisation are often forgotten. Driving traffic to a website via search engines is all well and good, but if your landing pages are not user friendly then all your SEO efforts will be in vain.
After all, would you buy something from a high street shop that looks messy and disorganised because you saw a great TV ad for them?
The principles are the same for a website, so it is vital to ensure that your pages attract visitors that convert and bring in revenue.

Conversion Optimisation?

But how can I optimise my pages I hear you ask....well, conversion optimisation and testing are the answers.
To put it simply, conversion optimisation is the art of making your website perform better (by improving conversion rates) whether you are after extra sales, quote requests, document downloads....and a lot of the optimisation will go through testing your pages and forms using A/B tests at first, and moving on to slightly more advanced multivariate tests (aka MVT). Like many other things online, conversion optimisation is not an exact science, hence the need for tests.

So to help you get started, here is a quick overview of what needs to be done to set up your first successful A/B test:
  • Set your website objectives
 The first thing you need to do is to establish a clear set of objectives for your website: what do you want users to do? Knowing what your objectives are will help you focus your efforts on the right areas and pages i.e. if you want to generate sales online, then your focus will be on the sales funnel and key pages involved in the path followed by visitors.
  • Get the right tools for the job
 One of the best free tools available for conducting optimisation tests is Google Website Optimizer (GWO), so I'd highly recommend you to register an account to get started.
A lot of agencies offer bespoke testing services, but at a (very high) cost, so using GWO should save you a lot of money. Alternatively, you can use Google accredited agencies such as ConversionWorks who will assist and guide you along the way. Having worked with them and their resident conversion genius Doug, I can only highly recommend them.
It is also strongly advised to use Google Analytics to get more in depth understanding of the results.
  • Identify problem areas
 Now that you've established your objectives and signed up for the testing tools, you're ready to analyse your website data. Looking at visitors' behaviour will very often highlight areas of concern or raise questions about the performance of key pages. If you're using Google Analytics you have probably set up some goals that match your objectives, so these should be your starting point. Are visitors dropping off on a specific page/form? is the bounce rate too high on a key page? Understanding your website data is key to testing.
  • Choose a test page and conversion point/goal
Once you've identified where potential problems lie, choose the page that you want to test and think about what could/should be improved to increase conversions on that page. For conversion optimisation purposes it is crucial to avoid assumptions and personal opinions. Remember that what matters is what users think and do, not what you think is best! A new design or colour may not look as nice as the one you originally had, but if it converts better, then your job is done.
To judge the success/failure of a test you also need to have a goal/conversion point to measure against. This could be a sale, a click through or whatever other metric is the most relevant for you. This goal will be added in GWO to assess the test results.
  • Design test pages
A typical A/B test will usually require you to test the original page (control page) vs a new one against a specific goal/conversion point. In order to do this you will need to create a different version of the test page based around what you think are the potential problems in the original e.g. if you think a button is not the right size or colour in the existing page, design the test page with a different size and colour. For A/B test it is better to test a completely new design, but it is up to you to decide how different the two versions will be.
A/B is just a start, but if need be you can test more than 2 versions of a page to run an A/B/C/D.....test. If you're a beginner it is better to start with a simple test and refine as you go along.
  • Set up the test
You're now ready to set up your test in GWO. While this is fairly straightforward, Google provides some very useful guidelines if needed. To start the test, you will be required to add some testing tags to the test pages including the conversion page, so having direct access to your site's code or a website developer will be useful.
You'll also have to decide what percentage of your site traffic should see the test. The higher the percentage, the quicker you should get results.
Once you have completed these tasks, put your pages live and wait the results to come!
  • Analyse the results
 Depending on how much traffic goes through your pages, results will appear more or less quickly in GWO, but useful graphs and conversion rates will help you understand how everything is performing. The GWO user interface is quite self explanatory but once a clear winner is found it will be clearly shown in green. All you need to do then is set the test winner as the version that should be displayed on your website. If it turns out that the original version is performing better that the new one, then keep it. At least you will have eliminated doubts you had about it.
  • Repeat and refine
As tests are often a trial and error process, repeat and refine tests as you go along to constantly improve onsite conversions. Designing a clear testing plan is recommended to keep activities focused.
  • Final word of advice
 Finally a few words of advice to test novices:
  • Avoid assumptions when designing tests, put users first, not personal views/opinions
  • Understand your analytics
  • Do not run too many tests at the same time so that you can clearly assess what is impacting conversions
  • Think continuous improvement rather than point in time changes
While this is not a fully comprehensive checklist, it should help you get started in the world of website testing and conversion optimisation.
As usual feel free to contact me if you have specific needs for your business.