Product recommendations can multiply profits. 

Unfortunately, not all recommendations are the same. We've foundpersonalized product recommendations vastly outperform generic ones. 

This guide walks through how to implement personalized recommendations in your store. Click here to skip to the examples, or read on for the full guide.

Note: this page was updated on September 7th, 2020 to reflect the latest findings on product recommendations, content personalization, and their effect on eCommerce sales. 

We covered in detail how advanced product recommendation engines work here.

However, in brief, I like how Amazon details how their recommendation engine works. 

Above, Amazon's recommendation engine incorporates a variety of inputs to create their personalized recommendaitons. 

Barilliance helps eCommerce store's create predictive, effective product recommendations with machine learning and AI capabilities.

Here is how it works. 

Personalization depends on customer data. 

Barilliance incorporates three major sources of data to create personalized product recommendations. 

They are:

1.Aggregated data (category/product views, adding to cart and purchase data, internal search queries, etc.)

2.User specific data that is used to personalize the recommendations. Similar to aggregated data, user data is the specific user interactions such as which categories and products the user viewed, bought, etc.

3.Static product data that is supplied by the client in the product feed. Product feed data typically includes price, availability, brand, tags, and other product attributes.

To create effect personal product recommendations, Barilliance uses a variety of machine learning optimized algorithms.

Our AI technology selects which algorithm to use to fill the product recommendation widget based on who the user is and in what context they are viewing your site. 

To illustrate, take the home page experience.  

The visitor could either be a new visitor or a returning visitor. 

If the user hasn't visited the site before, than a series of best selling products will be displayed. 

However, if the visitor is returning, visitors will see personalized recommendations based on their previous engagement with your brand such as:

-    Products related to their recently purchased items
-    Products related to their recently viewed products
-    Top sellers from their recently viewed categories

Above, our internal studies show that returning visitors have significantly higher add-to-cart rates compared to first time visitors. This, in part, is driven by content personalization including product recommendations that are more effective due to increased user information.  

Finally, you have the ability to define merchandizing rules for any number of demographic or behavioral segmentations.

We covered merchandizing rules in our article [Guide] Advanced Product Recommendation Tactics to 3x Revenue.


From there, we shared how:

"the best engines allow retailers to "overrule" the software's recommendations in lieu of explicit merchandizing rules you set up.

Examples include:

Above, Dearborn Denim highlights a series of complementary belts for the currently viewed wash of jeans.  Highlighting complementary products allows the customer to better envision how they will accessorize the item themselves, beyond what the product images display. 

Further, recommending complimentary products are a great way to increase the customers average order value by adding incremental revenue to the shopping session.

Buy Buy Baby provides another personalized product recommendation example. 

The first series of product recommendation widgets they show on the product display page bring forward similar products.  Here, instead of finding complimentary products based on the product's attributes, it surfaces similar options from the product catalogue. You can see that the first product recommendation widget shows similar pant styles and colors. 

The second product recommendation widget showcases a broader selection of alternatives based on the user's affinity. The affinity is determined in real time based on in-session shopping behavior. 

While pulling specific competing products helps deepen the user's search for the right product, Buy Buy Baby also broadens the search by recommending related categories and related searches.  

Above is a great example of recommending entire product categories and common searches. Using searches is a great way to focus on customer wants and help customers find what they are looking for. 

Amazon is a pioneer in personalized product recommendation widgets.  Still today they facilitate comparison shopping better than most eCommerce stores. 

Above they place user generated data (ratings), site data (in the form of dynamic product labels), and product feed data to let customers quickly compare similar products.

Above, Mint Julep showcases the main item (a printed dress) with accessories. They make it easy for customers to purchase the entire outfit with their personalized product recommendation widget titled "Shop the Look".

Order confirmation pages are an often under optimized step in the customer lifecycle. However, these pages represent a great opportunity to speak to customers at their highest level of engagement. 

Above, Amazon presents a series of recommendation widgets after they confirm the order details. In this case, the personalized recommendation widgets are titled "Items from our brands inspired by your views" 

Demographic data is an excellent source of customer information to personalize product recommendations. Simple examples include simply presenting the right categories based on which items are most applicable to the customers. 

However, my personal favorite example of using demographic data to personalize recommendations comes from Target. 

Here, they use customer demographics and past purchases to not only identify current needs, but also predict future ones. 

Below, Target gives a clear offer they know most families with babies eventually need - formula. Note how the reward is a gift card, which not only locks in a repeat purchase, but also helps establish Target as the place to get this staple.

Amazon provides another great example of using data to personalize offers. This time, they use date time. Recognizing when holidays and buying events are for your various customer groups is a great way to not only craft outreach campaigns, but also implement onsite content personalization strategies.

Below, Amazon uses an upcoming Holiday (Mother's Day" to create a series of personalized product recommendations. 

To demonstrate how effective personalized product recommendations are, we've gathered data on how recommendation widgets impact eCommerce stores across the customer journey. 

Below we look at statistics for: average order value, revenue, conversion rates, and shopping cart abandonment rates. 

Personalized product recommendations dramatically increase AOV (average order value). 

Sessions that do not have any engagement with recommendations have an AOV of $44.41. 

This number multiplies by 369% when prospects engage with a single recommendation. The effect continues to climb until tapering off around 5 clicks.

It is clear that the more personalized and engaging recommendations are, the more stores benefit from larger purchase orders. 

*Note: This study went across multiple industries. The significance of this study is not the nominal amount, but the relative increase. 

We conducted a study across 300 randomly selected customers. Here's what we found. 

Product recommendations account for up to 31% of eCommerce site revenues.

On average, customers saw 12% of their sales attributed to our product recommendation product. 

We also found that product recommendations increase conversion rates. 

Above, we see the conversion rate of sessions increase in lock-step with their engagement.  

Again, the biggest improvement occurs at the first click. Prospects who do not engage with recommendations convert at 1.02%. That number increases 288% after a single interaction. 

Our findings fell in line with a similar study conducted by SalesForce. They found shoppers that clicked on recommendations are 4.5x more likely to add items to cart, and 4.5x more likely to complete their purchase.

Lastly, recommendations have a significant effect on shopping cart abandonment. 

Here, we defined cart abandonment as sessions that completed a purchase divided by the total sessions that prospects added an item to their cart. We then segmented these numbers by how they engaged with recommendations in that session. 

We found that sessions that did not engage at all with recommendations, but simply added an item to their cart were much more likely to abandon their purchase. 

In fact, implementing personalized product recommendations can improve cart abandonment by up to 4.35%. 

Lastly, it is interesting to note that the effect on cart abandonment reverses after a certain level of engagement. This makes sense when you consider buyer behavior - especially those in the research phrase that use recommendations to find products.

Position of product recommendations influence how effective they are. We found widgets placed above the fold were almost twice as effective (1.7x) as widgets below the fold.  

Out of the 20+ product recommendations types that were reviewed in this study, the most engaging recommendation type  was ‘what customers ultimately buy’.

When a new visitor comes to your store, you don't know what products to recommend.

The best practice is to supply the best sellers of your store toward the top. You can also consider having multiple widgets, one for each of your top categories. 

As customers engage with your site, your product recommendation engine will begin to understand what types of products this customer is interested in, and supply more personalized suggestions.

Position of product recommendations influence how effective they are. We found widgets placed above the fold were almost twice as effective (1.7x) as widgets below the fold.  

This falls in line with our findings on dynamic content that increases conversion rate.

Another great way to personalize emails is via product injections. Software like Barilliance can inject product recommendations directly into the email.

The widget is tailored to reflect the products each customer is most interested in. Below is a great example of tailoring suggestions based on gender.

Below is an infographic we built with some of the key product recommendation stats we found. 

Product recommendations serve as the foundation for your eCommerce personalization strategy.

The next step to increase conversions is to build out more advanced personalization tactics.

Lastly, to see if Barilliance is the right product recommendation engine for you, schedule a brief demo with us.