5 steps to making a start with A/B testing
Through sharing and caring, trial and error, travel brands can improve conversions, as Mariam Sharp has been finding out
Also known as split testing, A/B tests help improve conversion rates. By creating two different versions of the same piece of content, you can gain hard data about what attracts and appeals most to your visitors and leads to conversions.
As mentioned yesterday, Expedia carried out 800 A/B tests on their online brand last year. At the Travel Distribution Summit, Europe, Isabelle Pinson, Senior Director Market Management, Northern Europe, Expedia, highlighted two possible examples.
Might one large picture on the front page be more effective than additional smaller pictures?
- Or a yellow button more effective than an orange button?
At the time of writing, a quick glance at the site shows that smaller pictures and a yellow button with a hint of orange appear to be the outcome of those two tests.
You might not have a budget the size of Expedia, but if you are not currently A/B testing here are five steps to making a start.
1. Focus in on your aim
What is the number one thing you need to measure with an A/B test? Without clarity about the main goal you want to achieve, analysing the results and using them to improve your site will be difficult. You need to know if you are meeting your objectives or not. See my earlier article - 5 ways to keep your eye on the ball with big data
2. Decide what feature to test
Like the earlier Expedia example, the most common subject online marketers test is the call-to-action, especially the button, that points web visitors to a sales page. Other aspects of the call-to-action are the banner or headline copy. You can also test the landing page; the thank you page; pay-per-click ads; email content, and social media traffic for example.
3. Be specific about the feature
The next step is to pick a variable of the feature for the experiment. Sticking with the call to action button as an example - variables you might consider including:
Colour - yes it can make a difference
Text - your content does count
Shape and/or Size
Picture vs No picture
Picture vs. Another picture
- Page placement
Only test one variable at a time! When you test more variables at once, it’s difficult to determine how each variable impacted performance.
4. Start the test
Now you have chosen your feature and the one variable to test, create two versions of the web page and begin testing. Give each version of your variable a different URL and then measure which receives the most conversions. For most experiments, you only need whatever analytics tools you already have in place to track your web traffic.
5. Analyse results
Gather a complete set of data. You need enough data to create statistically significant results before evaluating the outcome of your test. Once you have analysed results remember to share it widely across your teams.
To quote Daniel Waisberg, Analytics Advocate, Google: “Sharing is caring - it can help identify trends, rather than just data points.”
By working together you can focus on what the next important tests will be and then build this into a continuous practice. As Pascal Clement, Head of Travel Intelligence, Amadeus also said at the Summit: “Once you have an answer, you have another question…at the end of the day the story starts with data.”
Tell us in the comments how important A/B testing is to your business?
Guest columnist Mariam Sharp is a business consultant focusing on projects that promote international exchange. She can be found on LinkedIn and Twitter