Amazon’s Recommendation System Is Broken
Taylor CyganJuly 05th, 20184 minute read
Taylor is a content writing intern at Codal, authoring blog posts anywhere from UX design to other facets of the vast World Wide Web. Working alongside the talented members of the Codal team, Taylor works to produce relevant and engaging content. When she’s not immersed in the world of development and design services, Taylor spends her time: reading classic American literature, binge-watching a Netflix series, or ordering an excessively large iced coffee from her local Starbucks.
Amazon is one of the biggest players in the eCommerce industry with a monstrous consumer base, and a net profit of $178 billion just last year. One way that Amazon stays ahead of the curve is their integration of innovative technologies into their services. We have seen this in the past with their one-click shopping, automated marketing, use of robots and drones, and more.
Amazon was also ahead of the curve with their recommended feature.’ Released in 2008, Amazon open-sourced what gives them the information for their recommendations: its deep learning software, called ‘DSSTNE.’
As any loyal Amazonian will tell you, after you purchase something, Amazon is quick to follow-up with more suggestions. Or, on their site’s feed, Amazon will monitor what they view and give them similar products.
However, I’ve got news for you, dear reader— Amazon’s recommended products feature is broken. Well, not literally ‘broken,’ but figuratively. Amazon’s product recommendations feature is not up to par, often suggesting items that are unrelated or don’t seem to fit what was just ordered.
How can this be ramified? As a writer for a software development agency that specializes in eCommerce, I’m going to delve into their recommended products feature. Buckle up, let’s go travel into the Amazon.
A ‘recommendation’ for Amazon
All terrible jokes aside, Amazon’s recommended products feature does not provide an optimal experience for their users. To improve this, Amazon needs to fine-tune their product recommendation feature.
Amazon describes that they filter products based on, “...items you’ve purchased, items you’ve told us your own, items you’ve rated, and items you’ve told us you like.” Amazon bases their recommendations on their user’s site interaction or other similar user’s site engagement.
Now this seems all fine and dandy. However, Amazon suggests products that don’t seem to best fit a customer’s needs. For example, if I purchase curtains from Amazon, they typically suggest other curtains I can buy. In this scenario, why would I need more curtains, if I just purchased them?
Amazon should improve their recommendation system by only suggesting items that go hand-in-hand together. Extending the previous example, if I purchased curtains, I could need: rings for them to be hung on, a curtain rod, so on and so forth. They do this a little now with their “Frequently bought together” feature; however, I think in this scenario, I should be shown other household items.
Amazon’s recommendation system could take a page out of Netflix’s book when it comes to their show suggestions. Netflix’s recommendations take form in their, “Other movies you may enjoy,” or “Because you watched (insert show name here),” phrases. Netflix analyzes what you watched previously and feeds you predictive suggestions based on your viewing history.
Thus, Netflix is feeding you new and fresh suggestions. Their suggestions could even be shows that you would have never considered watching previously.
Or, as Netflix’s VP of Product Innovation Todd Yellin would describe it,
"The three legs of this stool would be Netflix members; taggers who understand everything about the content; and our machine learning algorithms that take all of the data and put things together."
With this metaphor, Netflix is monitoring user behavior, tagging content that is similar, and feeding users the content accordingly.
Amazon should be providing a similar experience in their product recommendation, they should be monitoring user behavior and flagging items that would be related. Therefore, providing users with a more tailored shopping experience.
What this boils down to for businesses everywhere
Amazon should alter their recommendation algorithm to better fit their customer’s needs. Their algorithm needs a major facelift in order to compete with the other recommendation systems on the market. Amazon needs to be providing their customers with recommended products that go well together and are different items than what was purchased.
This can be a key takeaway for any business that has a recommendation system. If your algorithm is not providing your shoppers from having an optimal shopping experience, it can be a hindrance to your eCommerce business’ success.
Like our dear friends at Amazon, you should figure out how you can alter your algorithm to provide your customers with helpful suggestions. If you finetune your recommendations, it can lead to more sales and higher customer satisfaction.
It’s best that your business learn from Amazon’s lacking recommendation system sooner, rather than later.
If your eCommerce business could benefit from having more experienced eCommerce advice, I suggest you get in contact with an eCommerce web design company. Not only can Codal help you improve your recommendation system, but your website as a whole.
Take the plunge with Codal, don’t waste your business’ potential with a flawed recommendation system like Amazon’s.