The Recommendation Engine Penetration
Why stir the pot of brain juice in deciding, when it can be achieved artificially?
In recent years, the smartphone has transfigured into the shape of our primary tool of distraction. The urge to check the latest posts on our social media every few minutes is undeniable. Searching for the next thing to purchase from the wishlist, grasping the best discounts and offers on the ’n’ number of E-Commerce applications has made the average online consumer turn into a shopaholic. ‘Which movie or web series to stream tonight?’ is one of the bigger decisions the online community had to take this year announcing the arrival of ‘The rise of the OTT’ phenomenon. Creating music playlists for workouts, car rides and meditations is not too unpopular either. So, what is common in all these digital activities across a buffet of platforms? What is the backbone that keeps the content-hungry community hooked to these platforms for significantly more time than that was intended initially?
The answer lies within the domain of Machine Learning — ‘Recommendation Engines’. To put it in simple words, a recommendation engine is a data filtering mechanism based on machine learning algorithms that work on customer ratings or preferences of products or services. Take Netflix for example. Based on the Collaborative Filtering technique, the User 3 in the diagram below would be showcased The Avengers and Murder Mystery in his recommendations section. The behaviour exhibited by similar users is leveraged here to come up with the recommendations.
Another classical example would be apparel purchasing applications. For instance, when a user buys a top, she would be then shown recommendations on which jeans goes along with the top. A set of similar products are also shown, from which the user can pick and chose i.e, Content-based filtering. A hybrid of Collaborative and Content-based filtering techniques are also common in numerous digital platforms like YouTube or Amazon. Research on online Recommender Systems done by a team from MIT showed around 10% increase in user’s willingness to purchase a product proportional to its star-based rating and around 80% of content viewed on Netflix was via personalized recommendations offered to their users.
In essence, the user is spoilt for choices and tends to keep using the applications excessively. Speaking of excess usage, music streaming platforms constantly keep track of the user’s taste in a specific genre and pop up similar tracks in the menu. The example of Instagram Explore feature can’t be stressed enough. Random posts out of nowhere are fed to the users based on their viewing activity. More often than not, this leads the users to follow more accounts and set the stage for revenue generation from ads that creep in occasionally during their limitless scrolling. Financial agencies are also on the prowl, recommending users customized Credit card plans based on the analysis of their spending behaviour. The underlying Big Data on which such processing is done is constantly growing fourfold in volumes. Occasional pruning of datasets is performed by these platforms to remain relevant in their user recommendations.
From a user-experience angle, the system of recommendations is definitely a value addition significantly reducing search time and improving content quality all the time. It could be described as a kind of spoon-feeding where users are constantly fed products or services that they may seem to like, with probabilities of positive responses increasing with growing usage. Clearly, the Recommendation Engines are learning and advancing in their decision-making prowess with each passing day. Ironically, humans are heading towards an information era where their choices are made by Machine Learning algorithms and their outcomes.
Why stir the pot of your brain juice in day to day decision making, when it can be achieved artificially? We live in strange times with stranger times to follow.