How Machine Learning Will Change Analytics

In business you have to look at digital usage measurements as chess pieces—what pieces are on the chessboard, how they move, and what movements they cause downstream, can impact future decisions.

That downstream view usually includes A/B and multivariate testing.  Optimization is typically deployed to improve website elements that impact user experience.  Changes in the optimization test platforms for Adobe and Google Analytics reflect the effect of machine learning on increasing the accuracy of test results and helping wmake better decisions.

Adobe announced Auto-Target, a machine learning protocol that automates personalization-focused testing of elements to determine the preferred individual experience with media.  The feature is included in Adobe Target, the testing and personalization platform.  With Auto-Target, marketers can select and automatically test various layouts, images, and texts that define not only presented offers, but also impact the larger customer experience.

Meanwhile, Google has announced Google Optimize, an A/B testing platform.  The solution was originally part of its premium Google Analytics 360 suite, but now regular Google Analytics users will be able to use the tool too.  The beta is available here: Roll-out is expected through the next month or so.

Multi-Armed Bandits

The new features introduce a coveted prize of automated optimization—a wider array of probability experiments.  One option that holds potential for accurately testing interactions with website, app, and digital media is based on the multi-armed bandit theory.  It is an approach that can account for multiple paths of interactions, then learn the best ones that contribute to the desired conversion outcome—a clicked button, a download, or an online purchase.

The multi-armed bandit theory is meant to reflect multiple choices with limited resources.  Marketers with an omnichannel marketing scenario and limited budget will certainly appreciate the scenario.

To understand the thought process behind multi-armed bandit theory, imagine yourself at the slot machines in a casino. Each machine has one lever arm, and you suspect that one or even several of the slot machines will hit a jackpot more frequently: So in playing these machines there is a “method” than yields the most money. But you only have a limited amount of money, and can pull only one arm at a time.  And you can’t pay and play all day.

So how do you learn which machines to play; how many times to play each machine; and in which order to play them; all in the shortest amount of time? “Mission: Impossible”?

It raises the point that marketers face today. Online optimization once meant testing an element change on a website, all for the purpose of establishing which change will lead to more conversions.  Today there are many influences on conversion—not least thanks to multiple devices and multiple channels.  The challenge can be best addressed in digital marketing optimization testing using machine learning techniques.

How Machine Learning Helps

Machine learning can help by quickly testing combinations of those influences through iterations that ultimately lead to the selection of elements most likely to create conversion activity. (In machine learning, an algorithm automatically corrects itself based on feedback.) Imagine not just testing beacon-triggered media, a mobile landing page, and buy button size individually, but testing combinations of these media quickly.  This revolutionizes optimization, encompassing analyses which go beyond a singular action like the click of a button.

Automated optimization test platforms arrive just as enterprises are seeking deep analysis methodologies to improve customer experience and personalization.  A joint Forbes-Sitecore whitepaper says that customer experience and personalization will be the top digital factors for attracting and retaining customers by 2020. Thus optimization testing is receiving a fresh spotlight.

I mentioned in my article about tidy data that marketers are under pressure to establish practices that can reveal contextual insights and accurate customer segments, especially with more diversified sources.  Optimization—such as that promised by the advancements from Adobe and Google—can help refine those insights, making marketing gambles less risky.

Related Posts