Build a Better Product Recommendation Engine
With Modern BI

Julian Alvarado

Sr. Content Marketing Manager at Sigma

Smart product recommendations help us find the items we need – or want – online. Most of us have also added recommended items that we hadn’t originally intended to buy to our shopping carts.

A product recommendation engine that accurately personalizes the products shown to a customer is an invaluable tool for retailers. It will reduce cart abandonment, increase the value of transactions, and create loyalty, increasing customer lifetime value.

 This is the third and final post of our 3-part series for retail merchandisers.

Read Improve Your Upsells and Cross-Sells with Market Basket Analysis and How Data Drives a Successful Product Mix Analysis to learn how data analytics helps retailers personalize offers, maximize revenue, and more.

What Is a Product Recommendation?

A product recommendation engine is an information filtering system driven by artificial intelligence (AI) algorithms that predict a given customer’s preferences. As a customer is shopping, the recommendation engine presents products and services that customers would likely be interested in based on past purchases, interests, behaviors, and other relevant factors. Examples include personalized recommendations for products on Amazon, music on Spotify, and movies on Netflix.

How Recommendations Help Attract and Retain Customers

Product recommendation engines help you better meet customer needs and desires, leading to revenue increases. It’s estimated that’s recommendation engine is responsible for 35% of its revenue! Product recommendations increase conversions and encourage repeat purchases in three ways:

Reduce abandoned carts and increase transaction size

The right eCommerce recommendation to the right shopper at the right point in the customer journey can help prompt customers to complete a purchase because they found exactly what they needed or wanted. It can also increase the number of items customers purchase as they add items to their cart that they didn’t originally intend to buy.

Drive repeat purchases

Personalization improves the customer experience since it makes it easy to find relevant products. A positive experience will continue to bring customers back for repeat purchases that significantly raise their lifetime value.

Grow brand loyalty and evangelism

Customers appreciate retailers that save them time, consistently providing what they need when they need it. Customers who have a positive experience are more likely to choose your company over competitors, write positive reviews, and recommend your company to their friends and family.

Why Product Recommendation Optimization Is Challenging

Truly personalized product recommendation is difficult because it requires real-time analysis of billions of data points across dozens of touchpoints. To get the most out of your product recommendation engine, you can’t merely present products based on aggregate data — you must get granular.

Here are three reasons that product recommendation optimization is so difficult:

  • The customer journey is a labyrinth of touchpointsBecause the customer journey now averages over 60 different touchpoints, it’s nearly impossible to identify key points of influence, opportunity, and risk without the right tools. This is especially true when multiplied by hundreds, thousands, or millions of customers across all stages of the customer journey. You must be able to quickly analyze this plethora of touchpoints en masse.
  • Vast amounts of data are being generated by various platformsMillions of customer data points are streaming in across PoS systems, loyalty programs, online payment processors, and dozens of other platforms. To wrangle this data, you need to bring it into your BI team’s cloud data platform and use a modern analytics tool that was built to empower users to connect and work with different types of data quickly and easily.
  • There are influencing factors outside of your controlCustomer needs and desires are also impacted by weather, holidays, cultural and religious considerations, and other external factors. You must include relevant factors in your recommendation engine to capture purchases driven by them. And it’s critical to be sure that these recommendations are accurate since errant recommendations can turn off, offend, or distract the visitor. Good product recommendation technology will help you ensure you’re considering relevant factors and capturing accurate data.

How to Make Product Recommendations Better

So how can you build a better recommendation system based on product purchase analysis and influencing factors? Three steps will help you optimize your product recommendations.

Build a Customer 360 — Identifying every touchpoint across the customer journey will help you develop rules for your recommendation engine that ensure the best possible recommendation for a customer’s current funnel stage. Here’s a step-by-step guide on building a customer 360 that gives you a clear view of the customer journey.

Create customer profiles — Based on your Customer 360, you can build out profiles with a definitive set of demographics, behaviors, and characteristics. These profiles will enable you to deliver more personalized recommendations at scale. Look at groups of customers with similar ages, zip codes, behaviors, etc., and recommend customers’ purchases to others in the group.

Integrate third-party datasets for better insights — Your own data provides an excellent foundation, but when you augment your data with third-party datasets, you get even more powerful insights. You can use external data to build product association rules based on outside factors that directly impact the individual shopper. The Snowflake Data Marketplace is an excellent resource for live, ready-to-query datasets.

Examples of Companies Doing Product Recommendations Well

Netflix and Spotify are both prime examples of companies that have optimized their product recommendations. Here’s an overview of each company’s recommendation engine.


 Netflix’s AI-driven algorithm looks at nuanced threads within the content rather than relying on broad genres to make its predictions. Each customer profile is based on activity data: what they watch, what they watch after, what they watch before, what they’ve watched recently and at what time of day, etc. That data combines with additional data aimed at understanding the content of shows watched. Netflix splits viewers into more than two thousand “taste groups.” These groups dictate the recommendations that pop up on each user’s screen, which genre rows are displayed, and how each row is ordered.


Spotify’s recommendations are created by an AI-driven system called Bandits for Recommendations as Treatments (BaRT). If a customer has been using Spotify for some time, it recommends songs based on previous listening activities. It also suggests new music that it determines the user will like. Spotify considers every user action, including listening history, songs skipped, playlists created, social media activity, and location.

Optimize Product Recommendations to Boost Revenue

A strong product recommendation engine will reduce cart abandonment, increase transaction value, and create customer loyalty. It’s one of the most reliable ways to consistently grow revenue over time. By leveraging your own data and external data, and by using modern business intelligence and analytics tools like Sigma designed to help you make the most of product optimization, you can experience these benefits.

For more on product optimization and other strategies retailers can use to improve profitability, read How Data Analytics Empowers Retail Merchandisers to Thrive Through Disruption.

And to learn more about how Sigma empowers leading retailers to maximize revenue by influencing every stage of the buyer journey visit our Sigma for retail page.

Ready to visualize your data for actionable insights?