By Srikant Sastri, Co-founder, Cryon Data
Ever wondered how some eCommerce portals know your preferred tastes? The secret lies in what is called a ‘recommendation engine’.
A recommendation engine is a system that helps people choose what they want, in an easy and efficient way by displaying recommendations based on past data. For instance, if you search for books on Amazon and look for Paulo Coelho’s ‘The Alchemist’, the next time you log into Amazon, it will give you recommendations to other books of Paulo Coelho. This is an example of working with past data. Similarly, if you buy a phone from Amazon, a tab called ‘frequently bought together’ or ‘people who bought this also bought’ option shows up. This is based on data from other similar customer’s buying trends. eCommerce portals see high potential in the implementation of recommendation engines from a sales and increased profits perspective.
A recommendation engine rests its inferences on explicit and implicit data collection. Explicit is based on the ratings or comments the product item may have received from customers during purchase, and implicit is based on the browsing patterns—time spent on each item on the website, combination of products brought together, analyzing previous history, among others.
Typically, there are three types of approaches for the recommendation engine: Collaborative filtering, content based, and hybrid system.
COLLABORATIVE FILTERING—PERSONALIZED VIEW
Collaborative filtering is the recommendation engine based on historic behavior or prior use. This can be based on the same person or on other people with similar choices/tastes. In simple terms, this method finds common ground between different sets of sample data with similar tastes to recommend options that one may have missed out. This is done using data analytical tools. The engine predicts more accurately as the volume of data increases. LinkedIn, Reddit, YouTube, Netflix, Last.fm—are a few examples of companies who use collaborative filtering engines.
CONTENT BASED—YOUR CHOICE, YOUR TASTE
Content based recommendation engine throws up suggestions based on a single user’s preference in the past. For example, a user reads a review about a book, searches for a book on the Internet, or rates the book. Based on these trends, content based filtering can recommend the same book or similar ones to the user. The recommendations in this case are based on single user behavior and not multiple. IMDB (Internet Movie Database) and Rotten Tomatoes are some of the many companies that use content based engines.
HYBRID BASED—AN EFFICIENT COMBINATION
A hybrid system combines both collaborative and content based systems together to increase efficiency of the recommendation system.
In conclusion, the future of eCommerce rests on the accuracy of recommendations, and it will only get better from here as just plain search will give way to guided choice.