I’m an avid online shopper. It’s convenient and simple, and best of all, I don’t have to deal with annoying store clerks. These are attributes that are absolutely necessary in my ideal shopping experience. Amazon.com is one of my shopping destinations of choice because of its wide selection of products. It’s like Walmart with a cooler name and actual customer service. The online juggernaut also gets into my head via typical big data fashion — it always tries to guess what I want. The keyword in that last statement, however,  is “tries.”

Amazon thinks it has me figured out as a shopper. The retail giant knows that if I click through to an item on a search query — an item like “Legend of Zelda” — even if I don’t ‘add to cart,’ I have shown interest in said item. Ideally for Amazon, the next time I visit, I would see, click, and buy one of these ‘recommended’ items related to “Legend of Zelda” that are scattered throughout my view as I browse the site. Although Amazon’s concept is spot on, it’s not quite right in execution.

I searched for “The Legend of Zelda,” and as we all know, “Zelda” and all its associated products are exclusively Nintendo products. Therefore, if I’m performing search queries for anything “Zelda” or “Mario” or “Metroid”-related, you’d think that Amazon would recommend Nintendo-specific products to me.

This isn’t the case with Amazon’s recommendations. On my Amazon homepage, I have an entire “recommended” section devoted to — you guessed it — Sony and PlayStation accessories and games. Not only is this a waste of valuable real estate, it’s a missed opportunity. If Amazon’s recommendation algorithms were effective, they would surmise my interest in Nintendo and show Nintendo accessories and games in my feed. PlayStation 3 games are absolutely no interest to me as a Wii U owner.

Now I was curious; would Amazon’s algorithms make similar incorrect assumptions outside the realm of the video game console wars? The conclusion was what I expected. I performed search queries for iPhone and Apple products, clearly showing Amazon that I have an interest in iPhones and Apple products. But, back on my homepage, I get a section dedicated exclusively to LG and Samsung phones. I have literally asked Amazon in all its infinite wisdom to show and bring me directly to iPhones, and the retailer shows me something it should know I don’t want. It’s no different than asking a sales associate where the iPhones are and that associate trying to up-sell LG and Samsung accessories — it does not make sense. In the brick-and-mortar retail world, this would be a lost customer.

There are countless instances of these faulty “product recommendations” not only on Amazon, but all over the e-commerce web space. In theory, product recommendations are a form of suggested selling. It’s the same as merchandising like-products next to each other on a store endcap. If you put too many random products near each other, or completely unrelated ones, you’ll confuse and lose that customer to the store with better merchandising. Need proof? Look at the in-store customer experience in T.J.Maxx and Marshalls versus the experience at Target or Best Buy.

Effective recommendations are a proven way to up-sell and increase conversion. To be effective, however, relevancy is absolutely necessary. SLI Systems recently announced the availability of a new product, SLI Learning Recommendations, which uses the power of SLI Learning Search data to provide contextually relevant recommendations. To learn more, take a look at our Learning Recommendations product page, or contact me at max.bunag@sli-systems.com.

Max Bunag, who clearly favors Nintendo over Sony, is an enterprise sales representative for SLI Systems.