Machine Learning Makes Standard A/B Testing Obsolete
Author: Randall Malcolm
Let’s take a moment to think out of the box.
Can machine learning really take the place of A/B testing? The short answer is “Yes”, but let’s dig in a little deeper and see how machine learning & artificial intelligence works to create an individualized user experience.
The goal of traditional A/B testing and multivariate testing in UX/UI optimization is to ultimately determine the single best website experience for an audience.
An audience is composed of thousands (or hopefully millions) of people from different locations, at different times, with different wants and desires.
Is there really a one-size-fits-all template?
No. If there were, there’d be no need for A/B testing.
Reduce the size of your audience to 1.
Treat each user personally. Determine the exact user experience that converts that exact user. We’re not the only ones who think like this.
Forbes recently released an article stating “Personalized Commerce is the Next Frontier.”
How: Machine learning. After determining which elements on a page can change, machine learning will systematically test those against each user.
For first-time visitors, we use a bottom-up approach. Users’ on-site behaviors, demographics and external circumstances (time of day, weather, news sentiment) are computed to match UX/UI to each profile.
For returning visitors, we add in the previous sessions’ data to better triangulate the perfect “converting” experience.
Real-time: Seamless deployment of personalization is just as important as the personalization itself. Let’s be honest: If the user can’t experience the magic, then you might as well not have it.
This means that each of the ecommerce tools needs to work in tandem. Luckily, nearly all of the tools we use are set up to do this. Google Analytics, FB pixel, your CRM… are all data-driven and easily accessible.