When Suggestion Engines Get It Wrong

3 Apr

What do Cheez-It crackers and eyeliner have in common?

adjacencies

From a friend’s Amazon shopping page

Very little, you’d think. And yet, a friend recently posted this picture to Facebook. As she tried to buy eyeliner on Amazon, the “frequently bought together” algorithm suggested she add Cheez-Its to her cart. All for the excellent price of $6.77!

Made you laugh, right? Because these two products are so unrelated, this “suggestion” seems totally off-base. But if you think about your in-store shopping habits and translate that to the internet- it starts to make more sense.

Physical retailers often merchandise associated items near each other. A classic case of this “adjacency” strategy is putting peanut butter and jelly on the same shelf. Next-level adjacency strategy is used to “suggest” items you might want to add into your cart more spontaneously. For example, Trader Joe’s sells packaged olives next to its wine display, because they want you to trigger your interest in buying olives to go with your drinks. Another adjacency strategy comes via promotional displays: for example, a s’mores bundle promotion I wrote about a few years ago.

When you’re shopping online, physical merchandising and browsing are replaced by dropdown menus, filters and algorithms. These algorithms try to connect your current behaviors, past behaviors, and what other people “like you” are doing. Algorithms are supposed to be helpful, but they can’t always distinguish between patterns, and purpose. When the algorithm notices a trend, it takes advantage of that information. And so, while the Amazon algorithm “knows” that a lot of people apparently buy eyeliner and crackers at the same time, it doesn’t know why. The connection it’s making isn’t completely logical, and the product it suggested to my friend  didn’t make sense within the context of her shopping purpose. She was shopping for makeup- not food. So seeing food pop up on the side felt nonsensical and out of place.

But wait: there’s one more physical shopping “truth” we need to consider. You may think about makeup as a discrete category, but you’re also very likely to buy it while shopping for other things. Think back to your last physical trip to a big box retailer like Target- what ended up in your cart? I always marvel at the seeming “randomness” of what goes into my cart at places like that, everything from flour to soap to greeting cards. But unless you’re on a very mission-driven shopping trip, or at a specialty store, you’re likely shopping for more than 1 thing at a time. And once you walk out of Target, your cart may very well contain the seemingly unlikely duo of makeup and food.

So what we’re seeing here is a clash between our physical shopping behaviors, and our online shopping behaviors. When we shop online, we’re often in “research mode,” looking for helpful information. We expect websites to know our frame of mind, and cultivate a relevant experience for us. We expect to see helpful reviews and product information about the category we’re focused on, not a different category we may start shopping for later.

Annies SaltinesWhen I tried to replicate my friend’s experience on my own account, I saw a lot of makeup suggestions- and then a suggestion for Annie’s Saltines. Which likely means that people do use online shopping to “stock up” on their essentials across category. And yet, from a merchandising perspective, we expect more from our online retailers. If they have our data, we want them to use it well. For these websites to properly influence our behavior, they need to take into account differences in our mindset as we go through the ecommerce journey- and not just suggest that “eyeliner + Cheez-Its = the perfect assortment”.

Advertisements

2 Responses to “When Suggestion Engines Get It Wrong”

Trackbacks/Pingbacks

  1. The Annual Report (vol. 4) | Culture Cookies - January 16, 2017

    […] When Suggestion Engines Get It Wrong […]

    Like

  2. But I Don’t Like That | Culture Cookies - June 11, 2017

    […] it. Websites and apps suggest content based on previous actions. Sometimes they mess up and use the wrong signals to assume interest. Maybe you watched Bring It On to wax nostalgic with a junior high friend, […]

    Like

What Do You Think?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: