Using AI, Predictive Analytics, and Recommendations

This is an overview of what a recommendation system in retail is and how we implemented it at a grocery chain.

These days, recommendation systems empower social networks, healthcare, finance, and e-commerce. At the end of 2016, Starbucks announced that they will be implementing an AI-based recommendation system in their cafes all over the world. This means that predictive analytics has finally found its way into retail.

Like e-commerce entrepreneurs, retailers can now send customers personalized offers based on their behavior. In other words, when you purchase your morning coffee, you’ll be automatically offered a fresh muffin. When you buy steaks for grilling, you will get a reminder offering mustard, ketchup, or whatever else you’ll need for a barbecue.

Let’s look at an average grocery shop customer who needs all items from his or her dinner list ticked off. At the end of the day, the customer realizes that the list is missing, and it seems impossible to recall all the items. Thanks to a supermarket app, the customer can make a list from a food catalog. Right after adding pasta to the list, the app recommends them to buy a Bolognese sauce. The customer grabs the sauce and wonders how they ever forgot about it. The recommendation works because someone in the past purchased Bolognese sauce right alongside pasta.

Here’s what retailers can get from using recommendation systems:

Grocery stores cannot analyze customer responses to the offered good in real time. However, the majority of retail chains have loyalty programs and a database of receipts. This data can be enough to make specific recommendations for customers.

There are several implementation solutions for a grocery store. It depends whether the grocery has an app and a website.

There are three main approaches to a recommendation system development: a content-based approach, collaborative filtering, and a hybrid approach.

When we developed a recommendation system for a big Russian grocery chain, we used collaborative filtering.