Spotify and Netflix have radically transformed the entertainment industry, and the way consumers interact with content. Their success lies in an algorithm that gives new effective recommendations to users.
Internet services such as Spotify and Netflix have radically transformed the entertainment industry, and the way consumers interact with content. Central to the success of these services lies an algorithm that gives new effective recommendations to users. Given data about the users’ previous preferences, it is able to recommend new products to these users. This can create user attachment to a service, or a boost in sales figures. For instance, 75% of the TV shows that users watch on Netflix are recommended to them via their powerful recommender system. This is also the reason behind the introduction of the well-known Netflix Prize, a competition in which the company offered $1million to the team whose algorithm offered at least a 10% improvement to their existing solution.
There are two popular approaches to implementing recommender systems. The first, called content-based filtering, recommends products that are similar in content to those that the customer has liked in the past. Although useful, these systems are entirely dependent on the ability to extract meaningful information from a product, which in the case of movies can be difficult. Also, recommendations will be confined to the scope of similar products, which usually will only represent a small subset of the customers’ interests. The second approach, called collaborative filtering, recommends products based solely on the preferences of other customers. It is assumed that if two customers share one interest, then it is likely that they share another. The main challenge with collaborative filtering approaches is that customers typically share their preferences for only a very small fraction of your product catalogue, which affects performance. To overcome the limitations of both approaches, it is common to implement a hybrid approach.
Recommender systems are an extremely flexible framework that can be adapted to a multitude of scenarios, to satisfy consumers’ increasing need for personalisation. One example is the online advertising business. Given that a large fraction of the revenue made by some of the big technology companies originates from online advertising, it is easy to see how the ability to improve the relevance of adverts can convert into a significant financial gain. Another prominent example is that of the e-commerce business. In particular, the giant retailer Amazon attributes 35% of its revenue to the use of recommender systems. By giving personalised recommendations, the company can help consumers navigate through the myriad of products and enhance the shopping experience. In addition, emails with product recommendations compels consumers to buy more through their platform. Other less well-known examples include OKCupid’s approach to recommend dates.
