A/B Testing Your Mobile App

A test long used by web developers to make sites more appealing or boost e-commerce sales is now emerging in the mobile app space.

A/B testing puts two versions of a website -- or a particular web page or some other website feature -- in front of visitors to gauge preferences. Feedback from such testing can prove valuable: Determining what site elements engage users can improve the user experience along with the sales conversation rate.

While A/B testing is a common practice for websites, the same cannot be said for mobile apps. However, market developments of late point to a greater availability of such services. PathMapp recently launched with the aim of bringing A/B testing to mobile apps. In addition, Swrve, which provides mobile A/B testing as part of its real-time feedback platform, raised $6.25 million this fall to fund the company’s expansion. [Disclosure: Intel Capital is among the investors; Intel is the sponsor of this content.]

Time for Testing?

Swrve targets the game app sector with its testing service. Hugh Reynolds, chief executive officer at Swrve, says some mobile developers have embraced mobile A/B testing, while others “need some encouragement to get down and dirty.”

But Reynolds believes it’s only a matter of time before the majority of game makers pursue A/B testing. “The widespread acceptance of the concept of A/B testing in the online world -- think Google AdWords -- means that most game developers understand that sooner or later they’ll have to join the party,” Reynolds says. He contends that mobile A/B testing will become commonplace in the game app space within a couple of years.

On the other hand, Convert.com, which provides A/B testing software for retailers and agencies, has discovered that customers, for the most part, aren’t quite ready for mobile app testing. The company in October disclosed plans to support users of Clutch.IO, a mobile A/B testing company acquired earlier this year by Twitter. Following the acquisition, Twitter said it would disband Clutch.IO’s hosted A/B testing service.

Convert.com sought to pick up the mobile testing slack, but found few takers. “It turned out that there was very little interest,” says Dennis van der Heijden, Convert.com’s chief executive officer and co-founder. “We thought it was a great opportunity, but our client base is more interested in testing the mobile versions of their sites than testing mobile apps.”

As it happens, the company’s interactive agency and e-commerce customers tend to outsource mobile app development to a third party, van der Heijden explains. So customers don’t become involved in mobile A/B testing, if any occurs.

A segment of Convert.com’s customers -- about 15 percent -- may look into testing their mobile websites, however. Leading e-commerce clients have expressed interest in a beta version of Convert.com’s mobile testing features, according to van der Heijden. The other 85 percent don’t have mobile websites. The need for mobile website A/B testing will increase as the percentage of e-commerce purchases made via mobile devices grows, van der Heijden suggests.

A Need for Testing

Game developers seem a bit more enthusiastic about mobile A/B testing than their e-commerce counterparts.

“As a mobile developer ourselves, we have a definite need for this that’s not being met by the current industry,” says Michael Orlando, chief executive officer of IDC Projects, a Rolla, Mo., research and development company focusing on games.

Orlando says he has heard of a number of places to go for in-app A/B testing. But he would like to be able to test outside of an app for external factors such as app names and icons.

“The importance of these factors is really high for those that want to have a big presence in the app store,” he notes. “The advertising needed in order to get up to the top of the stores relies heavily on these factors.”

Orlando says IDC Projects has created a tool the company uses internally to conduct some A/B testing inside its apps. The tool generates a custom HTML page that presents a player with two choices, which Orlando says could be anything from screenshots to general ideas for a game. The player then selects which one he or she thinks is better.

“Even if they choose randomly, if we do it enough, say over 1,000 responses, we can get a fairly accurate idea how good one idea is over the other,” Orlando says.

Orlando has not seen this type of testing in other apps or in any kind of commercialized format such as an SDK or a web service. He says the company may open the tool up to other developers, if there is sufficient interest in
by John Moore