Saturday, February 07, 2015

a/b testing

... One of Zulily’s core competencies is its content-management system, ZuStudio, which is designed to aggressively A/B test—or rather, A/B/C/D/E/F/G . . . test—everything you click, from the types of products that show up at the top of your website, to the kinds of photos that you and your family might identify with, to which photo angles might perform better than others.


WE’LL HAVE FOCUS GROUPS, AND MOMS WILL SAY, 'I’M YOUR PERFECT CUSTOMER, BECAUSE EVERYTHING YOU LAUNCH IS EXACTLY WHAT WE LIKE.' WELL, YEAH.

Zulily’s audience might not realize it—in fact, most don’t—but every consumer-facing experience, whether it’s on Zulily.com, on the mobile app, or through the daily email, is personally tailored down to a granular level, in real time. For example, if you have an 8-year-old boy, and you sometimes buy him clothes, Zulily’s algorithms won’t show you dresses for 3-year-old girls. What you see on your screen is engineered to be as close a proxy to you in real life as possible. This kind of one-to-one personalization is constantly being refined: Based on the way you browse, the software can infer the age and gender of your children; what kind of pets you have; what size you wear; and even whether or not you shy away from gluten. Like Amazon, the site is designed to feel intensely personal for millions of people. "We’ll have focus groups, and moms will say, "I’m your perfect customer, because everything you launch is exactly what we like,’" says Vadon. "Well, yeah."


These multivariate gains aren’t huge: Typically the change hovers between 0.2% to 0.4%. Yet those fractions of percentages add up, particularly when you consider the sheer amount of content Zulily publishes every day. It is fluidly adaptive. Like snowflakes, it’s never the same for any two customers.


Running millions of these tests every day is, in Zulily’s estimation, one of its core advantages in online retail, helping its buyers attenuate to the fickle needs and wants of moms as their children grow up. Children’s clothes, for example, are no longer Zulily’s biggest-selling category; now it’s womenswear. It has been an incremental evolution that would not have happened thanks to Zulily's enormous trove of buyer data. "The other week I was talking to a scientist about their research cycles and why they use flies for testing," says Cavens. "It’s because they multiply so fast! The life cycle is so short, that they can do a lot of testing."


http://ift.tt/16Mxh9S






from lizard's ghost http://ift.tt/1DNKgCx

No comments:

Post a comment