How to do Product Page Recommendations Right
A sizable majority of Americans — roughly 8 out of 10, according to Pew Research — shop online regularly. And those shoppers are going to make a big pivot in the next few years: Voice and image-activated search may make up 50 percent of searches in 2020.
What does that mean? It means that consumers are, more than ever, motivated and able to build a customer journey beyond text and text search. Visuals, especially, are now being used online in an organic way that mimics a customer’s discovery in a brick-and-mortar setting.
That is, ecommerce is finally catching up with the way that shoppers want to shop online, especially where visual AI is concerned, and that is especially true in similar items recommendations.
If you’re thinking, “But product page recommendations have been around forever!” think again. There’s a difference between just sticking some marginally relevant items on a product page and mimicking organic discovery, and that difference in recommendation is a difference to your bottom line.
Drive the thrill of the find
Product discovery is an important part of every online shopping experience. According to an Infosys study, shoppers experience that “thrill of a find” with sites that offer personalized and handpicked recommendations over those that don’t, and 96% of shoppers said they looked to retailers for new product recommendations.
Delighting customers with things they’ll like is not a new tactic — every ecommerce site with a “just for you” newsletter knows that.
What’s different about similar items driven by visual AI is that you’re integrating customized product discovery for every single person who visits a product page. You don’t need a return shopper, an email, a login, or back-end data scraping to figure out what a shopper will like.
Instead, you pull a recommendation from the product that the customer has visited. Whether it’s their first time or they’re your most loyal customer, they will see more of what they like.Say a customer visited this pair of pants on boohoo:
They can either scroll to the bottom for a “We think you’ll love” option or hit “Show Similar” to instantly find more sweaters like this one. Both features include the same results; here is the “Show Similar” pop-up:
The AI matches the pattern, color, high wasted fit and wide leg of the original pair of pants. Customers can compare prices and see similar styles to comparison shop instantly, without ever going back to search results.
Using the visual AI to power a similar items section on a product page makes the customer feel as though they’re going deeper on their shopping journey. With each item they click on, they’re being given a world of similar items to discover. While the section pages on ecommerce sites can be a jumble customers have to sift through, the similar items shopper is more akin to walking into a dressing room with a personal shopper.
Pair items with care
Visual AI isn’t the only way to recommend items to customers. It’s great for some sections of ecommerce, like clothing, but less appropriate for, say, an online kitchen store, where customers are not comparing hundreds of different stand mixers based on their aesthetic design.
For those types of ecommerce sites, using other kinds of recommendations on product pages is best. Similar prompts include “Customers who viewed this item also bought . . .” or “Frequently bought together.” These types of recommendations can be great — if they’re done with care.
A great example is Apple’s website. If you go to buy an iPad Pro, Apple tells you what you get with the iPad (charger, power adapter), and then recommends that you look at their keyboard and pencil, common accessories to the iPad:
Apple could even take it a step further and recommend an iPad case or headphones, and it would be appropriate because a customer buying a new iPad likely needs to purchase accessories for it. It’s a rare consumer who would have a latest-generation Smart Keyboard Folio lying around and no iPad to pair with it.
Where things can get sticky is when purchase intent is less clear.
Take Amazon’s recommendations for a bar of soap by Art of Sport:
Would someone purchasing a bar of soap be likely to purchase more soap? Or a bodywash? Maybe they’re comparison shopping, but if they’re searching for soap, they probably don’t want bodywash. Looking at four deodorants also seems like overkill, and some of those spaces might be better taken up with other items.
Infosys found that 77% percent of shoppers purchased an additional product based on a merchant’s personalized recommendations at least some of the time. To capitalize on that behavior, you need to understand the likely intent behind a purchase. While looking at aggregate customer data of what was bought together is a starting place, use analytics from your “Customers also bought . . .” section to home in on purchase intent.
If Amazon finds, for example, that people throw a deodorant in their cart but don’t buy the bodywash as much, they can swap that out for another product from the Art of Sport line, a razor, a body spray or some other personal-care product.
Of course, this is all speculation about Amazon’s customer, but it’s the type of investigation you should do on your recommended items before you compile some purchasing data and set it loose on recommendations unexamined.
Use social proof
A 2017 study by the University of Hawaii at Manoa that analyzed 27,175 recommendation prompts from five different online stores showed that product recommendations labeled as using social proof have a better click-through rate in various ecomerce categories than those with a generic “recommended for you” label:
- 43% CTR, “Others who viewed this, also viewed . . .”
- 35% CTR, “The most viewed . . .”
- 28% CTR, “Recommended for you…”
This shows that personalization and social proof drives shopping behavior. Shoppers, it seems, love seeing what items their peers are buying and take the word of their fellow shoppers to heart. Using this strategy can gently nudge customers toward items they’ll like because it gives those items a vote of confidence.
Take book retailer Barnes & Noble. You can’t really recommend books on a visual algorithm of how they look, and avid book shoppers tend to consult sites like LibraryThing and Goodreads to read reviews before purchasing. So it makes perfect sense that the retailer would add a “Customers who bought this also bought” section to their product pages.
Check out these recommendations off of the product page for fantasy book Six of Crows:
The recommendations are all in a similar genre and cater to those who like fantasy books with a twist of mystery, crime, thriller, or all three. This is a great way to help shoppers discover things they’ll like in a notoriously hard-to-recommend space.
Keep in mind the flip side of social proof, as well. The poorest-performing approaches from the 2017 study were generic calls to action:
- 6% CTR, “Picks for the month”
- 8% CTR, “Buy this”
This shows that generic messaging no longer speaks to shoppers — they know ecommerce companies can personalize and provide reviews, recommendations, and new discoveries that are tailored to each shopper. As the ecommerce world has exploded, the influencer industry has cascaded a wave of product experts, and competition has become stiffer, customers have demanded assurance that items are high-quality and bring satisfaction before spending their hard-earned dough.
Customer-first always wins
The thread between each of these best practices is simple: Put. The. Customer. First. When you conceive of your customer as a person on a journey to discover things that delight, it’s easy to see how you need to make your recommendations cater to your customers. Recommending items is about showing a customer you understand them and you want them to find things they truly love.