Whenever you see a recommendation for you on Amazon, we hope you know it’s not a coincidence! In fact, Amazon has become a case study for a lot of online retailers, to watch and learn about ways to focus heavily on data-driven marketing. At the crux of it, Amazon’s recommendation is based on many factors, like shopping history of a customer, the items customers select and store in the shopping carts online, items that they desire or like, basically their wish list, and what other customers have viewed and purchased. This mechanism is the brainchild of Amazon and they have christened it, “Item to item collaborative filtering” and it is this algorithm that Amazon has used aggressively to enhance, and customize the user experience, in turn targeting sales. Existing recommendation systems could not perform the huge volumes of Amazons products and customer base, hence they decided to build one for themselves. It is purely for this reason that a new mother would see baby products while a sports enthusiast would see gadgets on the Amazon web pages.
Using this methodology, Amazon has generated 102.1 billion in net revenue. 35% of Amazon.com’s revenue is generated by the recommendation engine.
Amazon’s recommendations are two types Online and Offsite.
Want to know how they do it? Well, read along……
Amazon’s On-line Recommendation, based on user shopping history and browsing activity for products, this list is created. The algorithm aims at the ‘wow’ experience. The sentiment, that only Amazon could have come up with this recommendation, and that if the customer was to search, they would never be able to discover the product, is targeted.
Frequently Bought Together, this feature only has one aim and that is to increase the order value. The customer might not need the product but buys them as they see others have. Amazon is able to show these suggestions to the customer, based on what they are about to buy, and cross-sell logical products that go with the primary one
Your Recently Used Items and Featured Recommendation, this is a very interesting recommendation tool that Amazon uses. Here, based on your browsing history, Amazon shows products with similarity, either different brand same product, same brand different products, products similar in shape, size etc…, with the intention of increasing sales by offering at least something that you find interesting enough to buy.
Off-Site Recommendations, Amazon sends specific and relevant emails to its customers to increase sales. This, of course, is done scientifically. A new mother would not receive hiking gear on sale updates in her inbox as it would be a disconnect. This form of recommendation has a very high conversion ratio.
Recommendations thus remain as one of the biggest innovations in online shopping. The prediction algorithms are very important for virtual stores, as their accuracy has a direct impact on sales. The data on which these algorithms work is updated regularly because the fluency in movement of a virtual customer is hugely based on their usage of the website, changes in preferences, introduction of new products etc…,
‘Collaborative’, ‘Customer Clusters’, ‘Simple Search’, ‘Item to Item’, ‘Bellkor’, are some of the most popular algorithms not only used by Amazon but other e-commerce companies for recommendations.
Predicting Algorithms used for recommendations, if they are fast and close to accuracy, then they highly benefit the sales of the organisation. Amazon’s case study is a testimony to it.
To learn more about the Amazon’s predicting Algorithm watch this space until next week for the big news!