Recommendation engines are surely a dream come true for many firms trying to provide targeted customer services across various domains. One of the popular examples worth mentioning is how Target, an American retailing corporation, figured out which of their customers is going to have a baby soon and further offered them with complimentary coupons to buy diapers1. Amazing as it sounds, this in fact was made possible by studying the consumption patterns which many customers leave behind through their store visits over the years. Similarly while making any item purchases through amazon, often you would notice lot of similar item recommendations which is nothing but the output from a recommendation engine set up to monitor your purchase activity, process the data using algorithms and then provide updates about your most likely next purchase. Another popular example would be about the success of Netflix, a movie rental firm, which provides accurate predictions about which movie you are going to watch next2. And by making the right recommendations for their customers, these firms in turn make more revenues through increased and repetitive sales especially from existing customers.
Also commonly known as recommender systems, in simple terms recommendation engines match the products or services that you might like with your personal preferences and needs. Let us take an example of recommender system that is used by an online book rental firm. Each book will have set of unique features such as title, author and genre. As a book reader, you have certain preferences towards a particular genre for instance thriller and say you have rated highly for Agatha Christie books. In order to figure out what should be the next book you might be interested to read, the recommender system scans through the entire database and lists out probably recommendations. And good part about this is these recommendations have a high probability of being among the thriller genre and most likely by the author Agatha Christie. Well you might think, this is straightforward but one thing about any recommender system is that the accuracy of predictions would increase depending on the frequency of interactions in terms of books you take and kind of information you share with the system. This additional information can be in the form of surveys or ratings you provided for other books and the recommender system will learn from these and over time would make better recommendations. While providing new recommendations, the system not only looks at your history but also considers book preferences of other users who are similar to you in terms of demographics and education background.
More and more firms are using recommendation engines to extract the inherent business value it has to offer. On a technical side, various machine learning algorithms that are used in recommendation engines are regression, decision trees, support vector machines, neural networks and others. And these are also the standard algorithms that are used to solve other data science problems related to forecasting and classification. Apart from the numerous advantages it has to offer for both business and customers, these recommender systems also pose a certain kind of security concerns. For instance, not all customers of Target would be happy about diaper offer coupons sent to them and this action can also look like a privacy invasion. In this era of data world, more and more people are also being concerned about the way different firms are trying to collect data about their customers like web browsing, purchase patterns or mobile usage sometimes without their knowledge or consent. I feel firms should create more awareness to the people about the kind of data they track and also how they use it to avoid losing trust on sensitive issues such as data privacy and security. Additionally they should also try to provide an option for their customers to opt out if anyone wishes to and further establish a fair balance in respecting their customer’s perceptions and selling more products or services.