Machine Learning in Collaborative Filtering and Personalized Pricing, a paradigm shift in E-commerce

As eCommerce has become the primary shopping tool for many, the data available on customer interaction has become seemingly infinite. While brick & mortar retailers are able to place lower priced impulse buys near checkout to incentivize customers to make one last purchase before exit, virtual storefronts have the unique ability to effectively make their entire catalog of products an impulse buy. These product recommendations account for roughly 31% of revenue in the eCommerce space, making the use of AI almost imperative to the industry. E-Commerce retailers giants can employ these technologies to drive product recommendations using their enormous store of data, and a major technique in ML known as Collaborative Filtering.

Collaborative filtering effectively uses the data from multiple customers who may share the same opinion and analyzes the purchases they make. This is effectively able to make the assumption that if two individuals are inclined to make the same purchase, then both individuals are inclined to have similar purchasing histories. Meaning if they were to get an add on with the product they were purchasing, then the other individual will be more inclined to do so as well (hence the “Customers also purchased” tab at checkout). These models are able to accurately place products customers may be interested in right in front of the user, something that retail stores couldn’t even imagine being able to accomplish.

AI and ML also allows for Dynamic Pricing, using advanced algorithms taking into account transient times, quantity of stock, and using predictive models to analyze which customers were likely to convert. Large retailers also bring in the utilization of price monitoring algorithms, which allow companies to maintain competitive pricing, closely monitoring other retailers. This rapid price flexibility allows for higher profit margins, as the calculations are able to encapsulate the moving variables of product delivery. While previously these calculations required constant supervision and attention, models are able to achieve these results faster and efficiently. Though dynamic pricing is able to take into account the various variables of the supply chain, true personalization comes via Personalized Pricing, using the predictive models mentioned previously to again predict individual customers' price sensitivity. These models take into account age, marital status, and shopping history to determine the willingness to convert on a certain product, and the maximum price they determine that they would be willing to pay.

While convenience in shopping may be the main drive for many customers, the reason eCommerce has been able to remain such a titan is very clearly the accurate recommendations, and rapid pricing strategies it is able to output. These strategies make it very clear as to why the conversion rates for those clicking on recommended products tend to be 5.5x higher than those who didn’t. Artificial Intelligence and Machine Learning have played a transformative role in this industry, and have not only driven these electronic storefronts, but have eased the lives for customers in the process.