Recommendation engines are the software applications that take customers’ shopping behavior and recommends them what other products they should consider buying. Recommendation engine plays two critical roles. First, it helps in personalizing customer experience on the web and in-turn makes online shopping a better experience. Second, it helps drive revenue for online stores by making sensible and relevant product suggestions to the customers. Amazon.com, the online retail giant, has spearheaded the innovation in this space. Everything from search results to the Amazon.com home page to emails sent out to customers are personalized for every customer with relevant recommendations based on the customer’s shopping pattern.
Amazon recommendation engine does wonders for the online bookstore. It understands a customer’s reading pattern from their search and buying history and recommends the customer similar books. The engine takes into consideration the book genre, author, buying pattern of the customer, buying pattern of other customers who bought similar books and a bunch of other criteria to display relevant recommendations. Over the years, Amazon has been able to sell millions of additional books and tap into its long tail with the help of the recommendation engine.
Now let’s talk about the scenario beyond books. Over the years, Amazon has expanded its retail footprint by selling products in more than three dozen categories. It sells everything from home appliances to jewelry. As expected, Amazon has adopted the recommendation engine for its other product categories as well. In many cases like movies, video games and music the adoption was very straightforward from books. With the help of relevant recommendations, Amazon is able to provide customers with a richer shopping experience in these categories.
Though in some other cases like home appliances, cellphones and gardening products, there’s a critical difference which makes the recommendations shown to the customers irrelevant. In categories like these, the online mega-store does not take into consideration the fact that if the customer has already bought the product from them, they won’t buy it again for sometime. I experience this when I bought a vacuum cleaner from Amazon last week. Even after buying the vacuum cleaner, my Amazon.com homepage has recommendations of vacuum cleaner and I am receiving email newsletter with attractive offers on vacuum cleaners. This would have made a lot of sense if I bought a marketing book and received recommendations for other marketing books, but when translated to a vacuum cleaner, this becomes an annoying experience. How can it be made better? Maybe by considering broader area of home appliances or cleaning products for recommendations than the narrow category of vacuum cleaners.
In all, the recommendation engine is a great innovation to enhance online shopping experience. But when applied to newer territories, there are lots of opportunities to make them smarter and more efficient.
The vacuum cleaner is a great example of how people relate to different types of products in very different ways. The limitation of horizontal recommendation engines is that they apply the same approach regardless.
In my opinion, what seems to be missing from Amazon recommendations are a notion of average lifespan of a product. Watches, washing machines, cars, etc all have far longer lifespans than books of videos.
Amazon should take this into account to make recommendations more relevant.
Lifespan is an attribute that can add some relevance. I think broadening the category can be very helpful as well.
I like how jinni.com does it for movies as well where they find recommendation based on a certain attribute for movies like mood or genre. Something like this can be applied to products as well where a certain attribute of the product is used for generating recommendation, like in case of vacuum cleaner, cleaning products can be an attribute, home appliances can be another one for generating recommendation.
I have been thinking about lifespan as an attribute for recommendation and think it can be really useful if applied in the broader algorithm. For examplei if a customer buys a cellphone with a two year plan, the store can start sending her recommendations for cellphone as her contract is near expiration (say from 23rd month onwards).
Similarly, if a customer bought a electronic razor from the store, may be in 8-12 months when the razor starts wearing out, he can get recommendations for accessories like blade replacement.
I think with the amount of information Amazon et al have about a customer, smart use of the information can really enhance the shopping experience further and drive revenues.
Pingback: Kindle application + e-bookstore = Advantage Amazon « Adscovery