The Scoop on Machine Learning for Optimization

Monica Knoblaugh

Jun 20, 2019


You’ve heard a lot about AI, Machine Learning, and maybe even Reinforcement Learning. But you may not know how these can be used to bolster your company’s growth initiatives. If you are not a data scientist you may believe this technology is out of reach.

If your company has a digital property, you need to be using Machine Learning. Top companies like Walmart and Amazon are already using it. Growth teams have discovered that machine learning is the secret to rapid iteration and a key driving force for growth. Traditionally, A/B testing would require long cycles to determine what works best for your audience on average. Even then, only one in every eight A/B tests would deliver conclusive results. These limitations to A/B testing generated a tremendous amount of wasted effort for growth teams. With Machine Learning, all segments of your audience are served up the variant they will best respond to. To better explain how Machine Learning for optimization works, let us use an ice cream analogy.

Imagine you are a first grade teacher. You want to bring the students ice cream on the last day before summer break. You ask the class, “How many of you like chocolate?” 40% of the students raise their hands. Next, you ask, “How many of you prefer vanilla?” 20% raise their hands. In an A/B test, this real-world example results in everyone being served chocolate ice cream. Chocolate caters to the majority making it the “winning” variant.

But, potentially 60% of the classroom may be left unhappy, not having received what they wanted. Instead of an A/B testing approach, what if we substituted a Machine Learning approach to solving this problem?

We would use context to better serve the audience.

In this classroom example, relevant context might exist in students’ behaviors in the school cafeteria. Have some students avoided dairy in their diets, perhaps due to lactose intolerance? Perhaps we could benefit from a dairy-free option if that is the case. Additionally, we could consider the number of students who failed to raise their hands as a signal that other flavors might be necessary to fulfill students’ preferences. We don’t live in a monochrome chocolate and vanilla world after all. With these basic changes to our approach, we could go from catering to 40% of the population to serving all of it. How sweet!

This was just a little sample, a taste of Machine Learning (AI).

Want to get the whole scoop? Watch our Free Webinar and discover how you can use AI for personalizing customer interactions in your organization.