Competition among online businesses is heating up with the digital transformation under the influence of the global pandemic. To thrive, e-merchants not only need to optimize their offerings but also have to apply different tactics to keep online customers on their site for longer and increase average order revenue.
Brick-and-mortar stores are heating up the online competition for e-Commerce with the lightning-fast digital transition.
One well-known strategy is online product recommendations. Similar to a store assistant in brick-and-mortar counterparts, eCommerce product recommendations offer a helping hand when site visitors are browsing around.
If you are new to online product recommendations, this article will give a perceptive analysis so you can apply a recommendation system on your eCommerce website easily.
What is an E-commerce product recommendation system?
Simply put, this is an automatic or manual online system to present relevant products across the customer journey online. Machine learning and customer data are the keys to selecting the right offering to showcase. A recommender system learns from a customer and recommends products that it finds most valuable among the available ones.
An E-commerce recommender system works as store staff, but customers don't need to ask for help or speak out their needs.
An E-commerce product recommendation system works in various marketing channels (website, social networks, emails) and across the discovery process (search engines, landing pages, product pages,...). However, in the scope of this blog post, we will focus on online product recommendations on e-commerce websites.
Why digital product recommendation is important
McKinsey has done in-depth research into the success story of Amazon and what’s a surprise is, “35% of what consumers purchase on Amazon come from product recommendations based on such algorithms”. That's a big number, which also proves the efficacy of an online recommender system.
Moreover, product recommendations help to build a personalized experience. Because they utilize customers' previous interactions and engagements to generate suggestions, recommended products are usually well matched to shoppers' preferences and interests. As a result, visitors keep coming back to the site, which increases loyalty and retention at scale. According to a survey report by Accenture, 58% of shoppers tend to buy from retailers that provide recommended options based on past purchases.
The 3Rs in customer loyalty and retention are Recognize, Remember, and Recommendations. Focusing on just one of the 3Rs can get more customers to come back. (Source: Accenture)
Another big bonus of an online recommender system is to nudge customers to buy more. A buyer, on average, spends 12% of their budget for recommended products, which accounts for up to 31% of e-commerce revenues in total. Obviously, product recommendations increase order revenues for online businesses.
How to recommend products when instant results not working.
How a product recommendation engine works in online stores
A recommender engine can make use of different algorithms and data to bring up the most relevant items to a particular user. It usually uses or combines some of the three following types.
This approach recommends products to customers based on a small set of items the customers have expressed interest in. The consumer basket here can be items from past purchases, items in the customers' wishlist, or products they have searched for or viewed. The system will pick similar merchandise to these items and recommend them to shoppers.
Any interaction of online visitors with the product can become a hint for recommendations. Therefore, it doesn't require deep customer insights and will be suitable for new visitors. Still, customers have to enter the desired syntactic product properties for the recommendations to appear.
Where & How to apply Attributes-based recommendations
An application of Attributes-based recommendations of Keyword-based recommendations on Homepage for returning visitors. It brings up merchandise related to previous search terms of the visitors. This way, users are guided toward items they’ve previously shown interest in. Also, searchers are highly convertible. Recommended products from search is a good idea to keep browsers on the right track of becoming buyers.
Amazon is the master in product recommendations on Homepage. They show “Recently viewed" items to remind customers of the previous shopping intention. (Source: Amazon)
Another best practice, especially for fashion e-Commerce is a “Shop the look" section on Product pages. It draws the shoppers’ attention to the whole set of outfits, a fantastic cross-selling tactic.
On the Red Dress website, when you view a product, they will suggest other accessories to “Complete the look". This saves customers from the dilemma of how to mix and match their outfits and is an effective cross-selling technique as well. (Source: Red Dress Boutique)
People-to-people correlated recommendations
In many research and articles, this approach is also known as User-User Collaborative Filtering. It mines the correlation between a customer and other lookalike customers (who have purchased from the E-commerce site) to generate product suggestions.
The algorithms of People-to-people correlated recommendations require a large customer pool so the suggestions are relevant and appropriate. It will take a lot of time and resources, which is not practical for all online businesses.
However, it is very effective. According to Google, “this allows for serendipitous recommendations”. Also, this recommender system can be trained automatically, without relying on the hand-engineering of features.
Where & How to apply People-to-people correlated recommendations
In practice, you can see a lot of eCommerce websites with the “People are looking for" section. It's an application of People-to-people correlated recommendations.
For first-timers, best-sellers, new arrivals, and trending collections are usually used as recommendations on Homepage because that is what people usually view when they first get to an online store.
It's a common practice for e-Commerce stores to display Best Sellers on Homepage. Katy Perry’s online store even puts the Best Sellers collection right in the center of the navigation bar, a super outstanding position. Camilla, on the other hand, makes a tweak in the microcopy by using “Most Loved”.
Homepage is also a gold mine to showcase social proof. You can use a section like “What's trending on Instagram” to display how your merchandise gets popular on social media. This can help visitors visualize the product in reality as well as encourage them to proactively share their purchases.
Instagram photos are of great help when it comes to recommending products in real-life use. Visitors can see how the items look, not on a well fit model, but on a casual person (Source: Mavi)
On Collection pages and Search Result pages, you are better off having best-selling sorting as default. Products that are selling like hotcakes will catch the eyes of many shoppers as well.
Showing hot sellers first can help increase sales for your top items. (Source: Red Dress Boutique)
User-User Collaborative Filtering in a recommender system is a well-done answer when a site search returns no hits. It applies to both the Instant search widget and No search result pages.
Don't let customers reach the dead-end of “No Products Found". Instead, show them some recommendations. Popular items will do a great job as shoppers like to catch up with the latest trend. Discounted products are also a wonderful option to keep customers staying and browsing around.
This is the most sophisticated among these 3 approaches. Based on users' profiles, a personalized system will recommend products that are similar to the ones that they have liked in the past.
Although this brings the most personalized experience that matches user preferences, it has many disadvantages.
Personalized recommendations require great effort from machine learning. It's like providing a personal assistant for each shopper. Furthermore, the recommendation pool becomes far smaller as it's built from one identity only. Sometimes, the recommended products are useless in initiating more purchases.
Where & How to apply Personalized recommendations
Personalized recommendations are an excellent tactic to increase order size. E-merchants can build algorithms to recommend complementary products for those already in the shopping cart. Added-to-cart items reveal the ultimate need of shoppers, so anything relevant to them is a potential.
Normally, e-merchants will show several product bundles at a lower price on Cart pages. This encourages e-shoppers to pay for the whole bundle, which looks like a saving, rather than an individual product. You can also suggest other similar products that are on promotion. When the customers see a better deal, they are likely to upsell the original item.
Just add some bread to the online cart and the recommender system of Yummy Bazaar will pick up some tea and marmalade so you can enjoy a wholesome breakfast.
To cut a long story short
An e-commerce product recommendation system is not easy to build, so you need to first understand how it works. After that, you can continue digging to find some apps or plug-ins to help with online product recommendations.
Boost Product Filter & Search is also working on this feature. Currently, we have released a recommendation feature for No search results on the Instant search widget and Search result pages. Follow us on social channels: Facebook, Twitter, LinkedIn for the latest app updates.
See you soon!