A / B Testing: The Most Powerful Way to Turn Clicks Into Customers
Format: PDF / Kindle (mobi) / ePub
How Your Business Can Use the Science That Helped Win the White House
The average conversion rate—the rate at which visitors convert into customers—across the web is only 2%. That means it's likely that 98% of visitors to your website won't end up converting into customers.
What's the solution? A/B testing.
A/B testing is the simple idea of showing several different versions of a web page to live traffic, and then measuring the effect each version has on visitors. Using A/B testing, companies can improve the effectiveness of their marketing and user experience and, in doing so, can sometimes double or triple their conversion rates. Testing has been fundamental in driving the success of Google, Amazon, Netflix, and other top tech companies. Even Barack Obama and Mitt Romney had dedicated teams A/B testing their campaign websites during the 2012 Presidential race.
In the past, marketing teams were unable to unleash the power of A/B testing because it required costly engineering and IT resources. Today, a new generation of technology that enables marketers to run A/B tests without depending on engineers is emerging and quickly becoming one of the most powerful tools for making data-driven decisions.
Authors Dan Siroker and Pete Koomen are cofounders of Optimizely, the leading A/B testing platform used by more than 5,000 organizations across the world. A/B Testing: The Most Powerful Way to Turn Clicks Into Customers offers best practices and lessons learned from more than 300,000 experiments run by Optimizely customers. You'll learn:
- What to test
- How to choose the testing solution that's right for your organization
- How to assemble an A/B testing dream team
- How to create personalized experiences for every visitor
- And much more
Marketers and web professionals will become obsolete if they don't embrace a data-driven approach to decision making. This book shows you how, no matter your technical expertise.
first, then personal information, then billing, and occupation/employer last. The optimized form yielded a 5 percent conversion increase over what had initially seemed to be the maximally optimized page. As Rush puts it: “You can get more users to the top of the mountain if you show them a gradual incline instead of a steep slope.” TL;DR More technologically or visually impressive pages don't necessarily lead to the user behavior you want. Experiment with keeping it simple and make any
several dollars per pageview, and this small change, together with our optimizations to the form and several other quick, simple tests, managed to bring in an additional million dollars of relief aid to Haiti. It's a testament to the power just one or two words can have. Find the Perfect Appeal: Wikipedia While the content on Wikipedia is the result of a vast collective effort, the Wikimedia Foundation is a team of just 157 committed employees who keep it all running behind the scenes.
of them. Try asking yourself: Is the language negative or positive ? Do you, for instance, advertise what a product has or what it doesn't have? Is the language loss-framed or gain-framed (e.g., the mammography study)? Is the language passive or action-oriented (e.g., LiveChat's “Try it free” button)? TL;DR There are endless word combinations to use on your website. Don't be afraid to brainstorm and think broadly: a testing platform lowers the “barrier to entry” of ideas, minimizes the
it: “Shifting the discussion from ‘What's testable?’ to ‘Everything is testable.’” Adopting the mantra “Always Be Testing” is one of the tenets of taking your testing program long-term. Redesigning Your Redesign: CareerBuilder and the Optimizely Website One of the mistakes we have seen companies make is undertaking a complete redesign of their site and then optimizing the new site using A/B testing. This is a violation of two core principles of A/B testing: define success metrics and
1900s). In most cases, we use the data we have to produce a test statistic. If that test statistic gets far enough away from 0, we conclude that the observed differences are likely not due to random chance. We will explore two such statistics. The Z Statistic for Two Proportions If our metric is a percentage or a proportion, the test statistic is given by: (E.6) From an earlier section, the denominator should look familiar to you. In its most basic form, it can be thought of as a measure