If you’re using a website to sell products or services, or to draw supporters to your cause, you want it to be the best it can be. Split, or A/B, testing provides the data you need to make better decisions about how to achieve this.
A/B testing is a simple experiment: You create two different versions of your website, randomly split your site traffic to each, then sit back and watch what happens. What happens depends on your goals. For most businesses, these goals are related to maximizing sales, customer engagement, ad spending, or optimizing product features.
How to Run Your A/B test
You don’t need to be a scientist or a programmer to run simple A/B tests. Here are the basics:
1. If needed, increase your website traffic
Since you’re splitting visitors into at least two test groups, you must have enough in each group to get statistically valid results. In experimental research lingo, this is your sample size. Ideally, your sample size should be in the thousands. If you’re not there yet, first focus your efforts on building traffic.
2. Decide what to test, and test one change at a time
To set up a successful experiment, you also need some reasonable hypotheses. If you already have good traffic analytics on your site, you probably have an idea of what visitors are doing there. These insights will help you decide where to start with testing. If you don’t have analytics, think about which tweaks might have the biggest impact your goals. Finally, only make one change per experiment; making more will reduce your ability to draw accurate conclusions from the test.
3. Ask IT for help
If you work for a company with a designated IT function and website administrator, ask these specialists to build you a separate experimental site on the website server and handle the behind-the-scenes random assignment of site visitors to the different versions during testing. They also should be able to provide you with result data and analytics.
4. Do it yourself
Run your own A/B tests by choosing from several cost-effective online split testing tools, such as Optimizely, Unbounce and Google Analytics Content Analytics. These tools use WYSIWYG (What You See Is What You Get) editors that allow you to create different versions of your site, run tests, and track results without needing any programming knowledge. WordPress, the popular open-source blogging and website platform, incorporates many of these and similar tools as extensible plugins.
How to Use Your A/B Testing Results
Your A/B test results will help you refine your site content to achieve specific goals, and may even lead you down potentially unexplored paths for improving your product features, pricing schemes, and marketing efforts. Here are a few ways you might use these insights:
Refine your existing product line. Perhaps customers consistently choose to buy one product when it’s recommended on another similar product’s page. Do you need both products? Are you cannibalizing your sales, confusing customers?
Narrow down the best pricing promotions and packages for certain customer segments and products. Your testing revealed more sales from individual products versus bundled options. Are customers eschewing them due to price? The data aren’t clear, so further testing seems in order.
Help gain a higher degree of confidence in where to spend your marketing dollars. Results can help you maximize ad spending, increase customer interaction with promotional materials, and allow you to make minimal copy upgrades that produce significant results.
Latest posts by Johannes Gehrke (see all)
- Simple A/B Testing for Results - June 3, 2013