For years now, many retail stores have gone through drastic technology transformations in order to stay on foot. Big brands such as Walmart, Target, Macy’s and Amazon have heavily invested in technology solutions to fulfill the constant needs the modern customer requires.

This is the first of a 4 part series of articles about technology innovation in retail. In this series, we will detail new trends and techniques being used by big players in the retail industry. You can check the other articles of this series in the links below once they are available. 

  • Part II: (coming soon)
  • Part III: (coming soon)
  • Part IV: (coming soon)

Machine Learning in Retail

Machine Learning (ML) is a set of artificial intelligence techniques that enable computers to learn how to perform tasks based on data. In retail, this technology is being applied in several ways in order to improve or automate processes. From CRM (Customer Relationship Management) to Recommendation Systems and Customer Value Analysis, ML techniques can be used to build custom solutions for all these problems.

The key point for using ML effectively is the availability of data in good quality and variety. Customer information, purchase history, volume of sales per product: all this information can be used to build models that allow us to discover in which demographic groups are our customers, how can we recommend products for them or which customers are more likely to stop coming back (churn prediction). By collecting data from mobile apps or sensors like cameras, we can predict which and when departments of a store are more visited or even how the visitors are reacting to the ads and items on shop windows. In the following sections, we list some ML applications in retail.

Recommendation Systems

Perhaps the most common use case for machine learning in retail, recommendation systems leverage customers’ choices, searches and past purchases to offer products that might be of their interest. Besides customers’ history, other kinds of data, such as demographic information, can be used to make more accurate recommendations.

The benefits recommendation systems bring to the table for both companies and customers are various – they make the shopping experience smoother, faster and more interesting. Customers like to be recommended products they would like. This put them in a better mood for purchasing products or services, improving the probability of a acquisition and improving the conversion rate. Recommendation systems are so common nowadays that most big retail companies use it in their e-commerce platforms.

Price Optimization

Price optimization is the process of offering prices that make customers more willing to pay while keeping the company gains. In other words, it tries to find the sweet spot between the customers’ and the company’s interest. This is done by analyzing not only the product’s production cost but also the acquisitive power of a typical customer, the location, the season and the competitors’ pricing. Making real-time changes to prices ensures the company always has competitive offers in the market. Companies that already use price optimization include Ahold Delhalize, Alibaba, Kohl’s and Otto.

Customer Sentiment Analysis

Sentiment Analysis is a popular topic in data science and, although it’s not a new technique, it has become cheaper and faster. Analysts can get data from both online feedback systems and social networks. In this case, data comprises of user’s comments about the products or brands, complaints, and suggestions obtained in several channels.

Using natural language processing, these texts can receive a sentiment score ranging from negative to positive. The texts can also be classified in compliments, complaints or other categories. This kind of analysis is invaluable when deciding how to improve services and products, measuring ROI of marketing campaigns or dealing with a crisis regarding the company’s brand.

Lifetime Value Prediction (CLV)

Customer Lifetime Value is a representation of the customer’s profit to the company over the entire customer-company relationship. The estimation of this value is made using all data up to the most recent purchase. By collecting user information such as preferences, behavior, demographics and recent purchases, we can model the relationship between this data and the customer lifetime value. Since attracting new customers is more expensive than keeping existing ones, CLV Prediction can be used to identify profitable customers faster and better utilize resources destined for the acquisition of new customers.

Conclusion

Machine Learning techniques, coupled with Big Data technologies, have given us the ability to assess massive amounts of data and extract value from it. Retailers are already understanding how important it is to utilize the techniques we detailed above and many others in order to stay in the competition. In the next article, we’ll talk about face recognition and how it is being applied in the retail business.

About the author

Matheus Gonzaga is a Data Scientist at Poatek. He loves learning new machine learning techniques and is interested in all topics involving AI. His hobbies are tabletop RPGs, videogames and climbing.