Every shopper knows what eCommerce product recommendations are -- even if they can't tell you exactly. They're the little carousels at the bottom or side of the page with headings like "frequently bought with" or "others also bought." These suggestions that online stores offer to customers for additional items they may be interested in purchasing have become a standard part of the online shopping journey, and are an important part of digital personalization efforts.
However, with the advent of artificial intelligence in retail, recommendations have evolved. Newer models can leverage a customer's specific purchase history and preferences to deliver personalized product recommendations that compliment and create an enjoyable customer experience. And by providing relevant, personalized product recommendations, eCommerce stores can increase their conversion rates and create a better overall customer experience, while also reaping the benefit of improved efficiency and back-end productivity gains.
Product recommendations are an important part of any eCommerce store’s customer journey and shopping experience. Studies have shown that personalized product recommendations can increase a store’s conversion rate by up to 4 times! This is because customers are more likely to buy items that they perceive to be relevant and useful to them. Product recommendations also make it easier for customers to find what they’re looking for, making their shopping experience smoother and more enjoyable.
In addition, product recommendations can help to build customer loyalty. By providing customers with a personalized experience and suggesting items that they may be interested in, stores can create a stronger connection with their customers and encourage them to keep coming back for more.
Product recommendations are an important part of any eCommerce store’s customer journey and shopping experience. Studies have shown that personalized product recommendations can increase a store’s conversion rate by up to 4 times! This is because customers are more likely to buy items that they perceive to be relevant and useful to them. Product recommendations also make it easier for customers to find what they’re looking for, making their shopping experience smoother and more enjoyable.
In addition, product recommendations can help to build customer loyalty. By providing customers with a personalized experience and suggesting items that they may be interested in, stores can create a stronger connection with their customers and encourage them to keep coming back for more.
The history of eCommerce product recommendations dates back to the early days of the internet. Initially, online stores would simply show customers a list of items that were popular, or related to what they had already looked at. These early recommendations were very limited, and relied on pre-programmed rules to determine which products to show. While originally novel, static recommendations like these required more manual intervention to maintain, and became outdated as personalization efforts advanced.
As technology advanced and stores began to use algorithms to analyze customer behavior, recommendations improved, suggesting more relevant items for customers to purchase. Recommendations started to leverage data on customer behavior (for example, purchase history and ratings) in order to deliver more personalized content. However, these engines ran into "cold start" issues, not knowing what products to show a new customer with no individual data, or when and where to show a new product with no associated ratings.
In recent years, these recommendations have become even more advanced, using machine learning and artificial intelligence to provide customers with personalized experiences. Today's market-leading solutions leverage AI-first solutions and advanced machine learning models, such Google's Recommendations AI, to deliver accurate and desirable recommendations even for niche and long-tail products. By leveraging the power of LLMs, recommendations can also react and respond appropriately to natural language interactions with customers for a highly satisfying customer experience.
This advanced AI can process user interactions in real-time, updating as a consumer navigates through the site for dynamic content updates. By integrating these recommendations across multiple channels such as web, mobile, and email, eCommerce retailers are able to provide seamless omnichannel experiences for their customers. These advancements have not only improved the shopping experience for consumers but also significantly impacted the profitability and competitiveness of e-commerce businesses.
The basic flow and function of eCommerce product recommendations is quite simple. The system collects data about a customer, such as their purchase history, browsing patterns, and demographic information. This data is then analyzed using advanced AI algorithms and sophisticated machine learning (ML) models to identify products that may be of interest to the customer. Products are then served and displayed on the website where the customer is interacting.
Aside from the flow, however, what makes eCommerce product recommendations "intelligent" is the magic of advanced machine learning models.
Machine Learning Models: The "Magic" Behind Recommendations
When it comes to eCommerce product recommendations, the "magic" behind the scenes is machine learning models. ML models are algorithms that analyze customer data and learn from it in order to make better recommendations. These models are constantly being tweaked and improved in order to provide customers with more personalized experiences and relevant product suggestions.
There are a variety of different ML models that can be used for eCommerce product recommendations, including content-based filtering, collaborative filtering, and hybrid models. Depending on the model used, different factors may be taken into account when making recommendations, such as customer preferences, past purchases, demographics, and more.
What is the difference between AI and ML?
Although at first glance they may seem similar, AI (Artificial Intelligence) and ML (Machine Learning) are two different concepts. And it's important that we stop here and talk about the differences. According to Google, AI can be defined as "the broader concept of enabling a machine or system to sense, reason, act, or adapt like a human." Machine learning, on the other hand, is defined as "an application of AI that allows machines to extract knowledge from data and learn from it autonomously."
Where AI and ML fundamentally differ is in their goals. The goal of AI, (again, according to Google) is to develop an intelligent system that can perform complex tasks like a human. In eCommerce product discovery, the best example of AI is in search. While search is essentially a retrieval function, understanding that a customer wants is essential to returning the relevant product results. The AI must interpret the search query, understand what the customer is looking for, and then match that request to items in the product catalog. It's not an easy task -- just ask any sales associate!
AI-first search and product discovery solutions have started using Large Language Models, which can be leveraged to understand user intent and context, for this application. This form of AI excels with long tail and natural language queries, providing a superior shopping experience for customers.
ML, on the other hand, has the goal of "building machines that can learn from data to increase the accuracy of the output." So ML, like the eCommerce recommendations models listed below, is more about using structured data to create self-learning algorithms to produce predictive models. Which is why it works so well for product recommendations -- recommending an add-on or complimentary product is best done using past customer data.
By leveraging the latest AI and ML technologies, eCommerce brands can provide customers with more personalized experiences and relevant product suggestions. This can help to build customer loyalty and encourage them to keep coming back for more.
Recommendations can be broadly classified into three categories:
- Relevant products: These are items that may be related to or complement what the customer has already looked at or purchased. For example, if a customer buys a laptop computer, they may also be shown printer cartridges and laptop bags as relevant product recommendations.
- Popular products: These are items that have been particularly popular among customers in the past. For example, if a store has sold a lot of electric guitars in the past, they may suggest an amplifier as a popular product recommendation.
- Personalized products: These are items that are tailored to each individual customer’s needs. This type of recommendation usually requires more advanced algorithms and data analysis tools such as machine learning and artificial intelligence. They also require clean and structured data, which can be a challenge for some retailers to provide. By taking into account more data points such as the customer’s demographic information, purchase history, browsing behavior, and more, the store can make more targeted and accurate product recommendations.
Each of these recommendation types will use and run different ML models, so eCommerce retailers should make sure they take advantage of all of them.
Now that you know what machine learning models are, you can start considering the different types that exist in eCommerce retail. As noted in the definition, ML models are all about using structured data to create predictive models. Different models will create different predictions and outcomes, and so there are several every eCommerce retailer should be using on their site.
To best serve your customers and provide truly personalized experiences, here are the 6 key ML models you should consider incorporating into your online shopping experience:
- Others You May Like
One of the most common -- and thus, most expected -- product recommendations models is "Others You May Like" which predicts the next product a customer is likely to engage with. This model specifically leverages shopping and viewing history, and greatly benefits from real-time updates to adjust recommendations as the user browses.
- Frequently Bought Together
Frequently Bought Together is exactly what it sounds like -- products that are purchased together within a single shopping session. This model is especially useful for creating upsell and add-on opportunities, for example when a customer has added a specific product to their cart, indicating purchase intent.
- Recommended for You
This model predicts the next product that a user is most likely to engage with or purchase. While it's most commonly seen on home pages, it can also be leveraged across category pages to suggest specific products within a given category for a customer.
- Similar Items
The Similar Items model is what most people think of when they think of eCommerce product recommendations,. This model presents other products that have mostly similar attributes to the product being considered. For example, if a customer is shopping for jeans, this slider would display different styles of jeans. It only requires product catalog data -- not user data -- and as such as one of the earliest models in use.
- Buy it Again
Ever bought something and been told you should "buy it again?" Yes, that is also a recommendation model. This model encourages purchasing items again based on previous recurring purchases. Using predictive modeling, customers are shown refillable or re-purchasable products that have likely run out.
- On-sale
Most typically used on the home page, the on-sale model type recommends on-sale products to encourage users to purchase discounted items.
Now that you've got your ML models, the other key aspect of product recommendations is their placement. After all, a customer cannot purchase a recommended product if they can't see the carousel!
The best way to ensure your customers see the appropriate product recommendations is to place them in key locations within your eCommerce website, such as at checkout or when customers are browsing a particular product category.
Here are all of the most important and prominent places you should have your product recommendations:
- Homepage
Most customers hit an eCommerce site's home page first, so it should be the first place you consider for product recommendations. The goal should be to provide customers with a personalized and relevant shopping experience, but since that isn't always possible right off the bat with new users, this is a great spot for category recommendations, popular product displays, and on-sale carousels.
- Search page
As discussed below, recommendations have a key role to place on the search page -- especially when a customer's query returns "no results!" When it comes to product recommendations on the search page of an eCommerce store, the recommendations should be second to the search results. Recommendations carousels often get placed at the bottom of the page, under search results, in order to provide customers with relevant and personalized results that match their search query. This is a great place for previously viewed products or recommended for you models.
- Category pages
When it comes to product recommendations on the category page of an eCommerce store, they can do several things. Recommended for you models can do well to highlight specific products in the category which are aligned to an individual customer's preferences and on-sale carousels can display discounted products from within that category. Within category pages it's important to remember that customers have often browsed their way to that page, and want to see products from that specific category. For a consistent shopping experience the recommendations should align with the category pages' products.
- Product pages
Product recommendations are also important on product pages, as they help to provide customers with more information about the items they are interested in. By providing complementary products or additional information related to the item, customers can get a better understanding of the product and decide if it is right for them. They can also complete a purchase if other components for their desired product are needed.
- Checkout/cart page
At checkout, product recommendations can increase the likelihood of customers making additional purchases. By displaying relevant items that could complement or complete their purchase, eCommerce stores can encourage customers to make additional purchases and drive up their revenue. Furthermore, customer loyalty programs can also be used in conjunction with product recommendations to incentivize customers to make multiple purchases.
- 404 pages
404 pages and null search pages have a lot in common. Typically, they're a dead end in the customer journey. As such, product recommendations here can play the same role as they do on null search pages. They encourage customers to keep going and look at other products they might like, and help recover lost sales.
Leveraging product recommendations in these six key locations will help to make sure that customers see relevant and interesting products at every stage of the online shopping journey that could lead to additional purchases. It is also important to keep in mind the timing of product recommendations, as customers may not be interested in making additional purchases if the suggestions come too soon after their initial purchase. As such, these strategic locations are even more important for boosting sales and revenue.
Now that we've covered what eCommerce product recommendations are and how they work, let's look at the benefits they provide, starting with the most common use case: out of stock scenarios.
Using Product Recommendations for ‘Out of Stock' Scenarios
eCommerce product recommendations can be strategically leveraged to address "out of stock" scenarios, turning what might initially seem like a drawback into an opportunity to enhance the customer experience and drive sales. 53% of consumers will abandon their cart entirely if there's even a single item they cannot find. Instead of becoming frustrated and leaving, product recommendations can help mitigate this frustration by suggesting relevant alternatives in the form of similar products or complementary items. This not only allows customers to find a suitable alternative quickly, but also keeps them engaged and loyal to the store.
Here are some ways in which recommendations can help mitigate the impact of out-of-stock products:
- Product Substitution Recommendations: When a product is out of stock, recommendations can suggest alternative or substitute products that are similar in nature, style, or purpose. For example, if a particular brand of running shoes is unavailable, the system can recommend similar running shoe brands or models with comparable features and price ranges.
- Cross-Sell and Upsell Recommendations: While a customer may be disappointed that their desired product is out of stock, recommendations can divert their attention to complementary or higher-end products. For instance, if a specific camera model is unavailable, the system can suggest related accessories, lenses, or even more advanced camera models.
- Pre-Order or Backorder Recommendations: Instead of losing a potential sale entirely, recommendations can encourage customers to pre-order or backorder the out-of-stock item. Highlighting estimated restock dates and offering incentives like discounts or free shipping for pre-orders can motivate customers to wait.
Other Benefits of Product Recommendations
- They streamline product discovery
When a product is out of stock, recommendations can suggest alternative or substitute products that are similar in nature, style, or purpose. For example, if a particular brand of running shoes is unavailable, the system can recommend similar running shoe brands or models with comparable features and price ranges.
- They improve customer experience
By providing personalized product recommendations, eCommerce stores can give customers a unique shopping experience that's tailored to their interests. In fact, studies have shown that customers are 80% more likely to purchase when an eCommerce business provides them with a personalized shopping experience. This helps to create an engaging and personalized shopping journey that keeps customers coming back for more. It also builds a stronger connection between the store and its customers, which leads to higher customer loyalty and satisfaction.
- They increase conversion rate
Relevant product recommendations reduce the time and effort customers need to find items they want. By displaying relevant product suggestions in key locations, customers are more likely to find what they need and make an additional purchase. This convenience can lead to quicker purchasing decisions and higher conversion rates. Additionally, the sophisticated ML models used in modern product recommendation engines can be optimized to increase conversion rate. Using customer and site data, these models can predict and display products that individual customers are most likely to purchase, boosting site conversion rate as a whole.
- They increase average order value
Similar to conversion rate, product recommendations can also boost the very important eCommerce metric: average order value (AOV). Product recommendations are a great way to encourage customers to make additional purchases and increase the overall value of their orders. Presenting customers with complimentary product recommendations is an easy way to do this, and ML models can accurately predict which items are most likely to add to cart -- as well as which are most profitable.
- They encourage brand loyalty and retention
Effective recommendations keep customers engaged with the platform, reducing the likelihood of them seeking alternatives elsewhere. Satisfied customers are also more likely to return to the eCommerce site for future shopping needs and they typically spend more than new customers. Thus, product recommendations often result in a higher customer lifetime value and increased revenue from repeat business.
- They increase revenue
In addition to boosting key site metrics like conversion rate or AOV, product recommendations are an effective way to boost revenue and increase the bottom line for eCommerce stores. Recommendations often include items that complement what customers are currently viewing or have in their shopping cart, increasing cross-selling opportunities, and can promote higher-end or premium versions of products to increase upsell opportunities. All of these options, of course, increase overall revenue and profitability.
Overall better user experience
Overall, the convenience of easily finding relevant products contributes to an overall positive user experience. As per Forrester's 2023 Customer Experience Index (CX Index), the companies that are ranking well are companies who deliver positive emotional experiences. Specifically customers want to feel happy, valued and appreciated.
eCommerce product recommendations are an easy way to deliver a positive shopping experience, delivering personalized recommendations at appropriate stages of the online customer journey to assist customers in their mission. In this way, eCommerce stores can give customers a unique shopping experience that's tailored to their interests. This helps create an engaging and enjoyable shopping journey that keeps customers coming back for more. With the right product recommendation engine in place, customers can easily find what they need and discover new items that they may not have considered before.
We've mentioned search several times so far, but what, exactly, is the relationship between product recommendations and eCommerce site search?
Well, the relationship between product recommendations and search is an important one. Product recommendations can be used to complement search results and provide customers with more relevant and personalized results. However, search results are inherently different than product recommendations. As they relate to the customer, product recommendations are passive -- they update dynamically and appear in response to a customer's actions, yes. But the customer does not have to do anything to actively receive product recommendations.
Search results, on the other hand, are inherently active. Customers are actively looking for specific products when they search, and display high purchase intent. 69% of all customers actively use the search bar to look for products, making it the most common method of eCommerce product discovery.
Where search and recommendations overlap is in the UI (though hopefully not literally) and in data. Customer data from search, including products a customer clicked on, can be used to tune and train recommendations models, further enabling them to deliver personalized recommendations. Recommendations are more about guiding users to discover products they might not have actively searched for.
Search and recommendations have a reciprocal relationship. Each supplements the other and helps guide customers to their desired products in different ways, but with the same effect: increased revenue, sales, customer loyalty and customer satisfaction.
When choosing an eCommerce recommendations solution, there are a few key factors you should take into consideration:
Scalability
Your recommendations should be able to scale with your business, effortlessly handling the nuances behind customer behavior. This helps to drive engagement across channels and deliver omnichannel shopping experiences as well.
State-of-the-art ML & AI
As discussed above, ML models are the heart of effective eCommerce product recommendations. Leveraging next-generation AI and ML is essential to keep up with rising eCommerce customer expectations. Avoid systems that run on legacy technology, as their limitations can hold you back from creating an excellent customer experience.
Optimized results
Your recommendations should be able to leverage user interactions and ranking models to meet specific business goals. You should also be able to optimize those models for specific business outcomes, such as conversion rate, click through rate, or revenue per session.
Security, privacy and compliance
Since recommendations rely on customer data, your solution must comply with security and privacy regulations. Compliance with the General Data Protection Regulation (GDPR) at minimum is essential.
eCommerce product recommendations are a powerful tool used to increase customer engagement and sales. Leveraging a wide variety of ML models allows retailers to display popular products that match a customer's interests, or suggest complementary items that could go along with their previous purchases. This helps ecommerce retailers leave customers feeling happy, valued and appreciated -- the three key emotions Forrester has identified as being critical to providing a great customer experience.
Strategic placement throughout a retailer's website allows brands to offer up personalized and relevant recommendations at strategic points throughout the online customer journey. These personalized suggestions are based on a customer’s purchase history, preferences, and other factors in order to make the shopping experience more enjoyable and efficient. And by tailoring recommendations to customers’ interests and needs, eCommerce businesses keep customers loyal and build relationships with them -- ultimately boosting sales and revenue.
Ready to upgrade your eCommerce recommendations? GroupBy's eCommerce Search and Product Discovery Platform powered by Google Cloud Vertex AI Search for Retail delivers next-generation Google-quality recommendations that drive customer engagement, loyalty, and revenue. Harness the power of true AI to deliver engaging product recommendations at every customer touchpoint and create a frictionless buyer experience!
