Amazon has in the past claimed that 35% of their revenue comes from purchases customers have found through recommendations. This is a true testament to the power of suggestion. In spite of this, product recommendations are an often overlooked part of the eCommerce shop experience. Embedding product suggestions on product detail pages (PDP) are an excellent way of creating new revenue channels, boosting upsell and cross-sell opportunities and helping the overall conversion rate.
Most of the biggest names in fashion already offer recommendations on their PDP. However some of them aren’t yet fully optimized for conversion. And there are still some who don’t offer any kind of recommendation system. We did some research into what type of recommendations the big eCommerce players offer and where exactly on the page they place them. After presenting our findings, we will demonstrate how Fashwell’s image-based recommendation engines serve brands and shoppers in the best way possible.
Best Practice on Recommendations at Zalando, Nordstrom, Amazon and more
To get a better picture of what kind recommendation systems already exist, we had a look at the sites of the top dogs in eCommerce. Specifically, we were looking not only at the types of recommendations they offer (i.e. how the recommended products relate to the main product), but also the placement of the recommendations and how many different kinds were offered. We were interested in the quality as well as the intent behind the recommendations. Here are our findings.
Above we have the best practice recommendation pages for Zalando (left), Nordstrom (middle) and Amazon (right). For both Zalando and Nordstrom, the general layout of the PDP is the same. Below the featured product first comes a “Shop the Look” recommendation section. The very top placement of Outfit Recommendations suggests that how to style a product is a priority for shoppers. Retailers are doing themselves a huge favor by offering this feature. It is a very effective way of upselling complementary products.
Next follows a “You May Also Like” and “Others Also Bought” section. Although these can be helpful to the shopper in some cases, they usually consist of personalized product recommendations that are based on clicks, views and popularity. In the case of Zalando and Nordstrom, they show visually similar products.
Looking at Amazon’s recommendations, we see they focus a lot on sponsored products and the recommendations appear to be mostly quite random. Out of all of them, Amazon has the most recommendation sections, which are distributed throughout the entire PDP. There are recommendations that even appear after scrolling through customer reviews, including suggestions of site-wide bestsellers.
Layouts of eCommerce Product Detail Pages
Below are the layouts of four of the top online fashion vendors.
Based on our research, most eCommerce players offer some kind of recommendation engine. However not all of them are optimized to take full advantage of the PDP. Out of context recommendations do not really serve the shopper – and this can mean a lost opportunity for conversion for the retailer. Although generally clicks on recommendations comprise only 7% of visits, they drive up to 24% of orders and 26% of revenue. Not to mention 77% of digital natives have come to expect relevant product recommendations when shopping online. This means it’s important for retailers to be even smarter about their recommendation engines.
Relevant Product Recommendations Through Image Recognition
Recommendation engines can be based on a variety of factors. A lot of them rely on user behavior, showing products that are clicked on or viewed the most, or best-sellers across the entire site. In the same vein, personalized recommendations tailored specifically to the user profile also exist. But seeing how collecting user data has become a prickly topic of late, there is an easier and even more effective way of providing tailored product suggestions: Contrary to other recommendation engines, Fashwell relies on image analysis to present relevant Similar Product Recommendations. We have found that recommendations based on images return the most relevant results for shopper. They can increase overall site conversion by up to 35%.
Our machine learning based recommendation engine understands the style of a product. This is how it recommends visually similar items from your product feed. You can increase the chances of conversion by offering shoppers alternatives to a product they’re already looking at. Perhaps they’ve clicked on a black high heeled pump, but they prefer the slightly rounder toe on the one in the recommended section. Similar Product Recommendations are incredibly useful for automatically refilling out of stock items. A sold out sign? Here is a selection of very similar ones that the shopper may also want to consider. These tools ensure that the customer doesn’t ever hit a dead end.
Fashwell’s Similar Product Recommendations lead to higher conversion rates across all product categories, and our customer Nelly have reported 250% higher engagement compared to their previous, non-image based and manually curated recommendation engine.
Automatic Shop The Look Outfit Recommendations
In addition, Fashwell offers Outfit Recommendation. This is an automatic outfit assembly tool that recommends products based on the full look a model is wearing on an image. Let’s say a shopper clicks on a pair of jeans, and sees the model also modelling a cool shirt and shoes. Fashwell’s Outfit Recommendation auto-populates those additional products on the same page to show shoppers “how to wear” the look. By embedding these product suggestions, brands can take full advantage of their images and offer several upsell opportunities. Outfit Recommendations increase AOV and eliminate manual work for visual eCommerce merchandisers.