How to Create Personalized Product Recommendations with AI

  • 7 min read

Personalized product recommendations are created by strategically showing relevant products to shoppers as they browse your website. Here are examples on how to use them:

  1. Show recently viewed items to your customers on product pages and checkout pages
  2. Add recommendations that relate to products users usually engage with; ‘You might also like’, ‘Complete the look’
  3. Generate recommendations that bundle products together or offer a discount when additional items are purchased; ‘Frequently bought together’
  4. Send out email product recommendations that can feature all of the above

Choosing the right recommendation engine, determining optimal placements and the number of recommendations, and implementing A/B testing are all essential steps in finding success, which we explore in this article. 

What are personalized product recommendations?

Personalized product recommendations in ecommerce are a way of showing relevant products to consumers as they browse your website, amongst other channels such as email, messaging and mobile apps. In some cases it’s just one product, however the recommendations could include a whole selection of items that shoppers would find appealing. 

The key to this ecommerce strategy is ensuring the recommended products are unique to individuals, tailored to them based on interests and online behavior. This is what makes the recommendations personalized and therefore an effective tool for marketers. 

 

Optimize your product feeds in minutes with AI

Types of product recommendation engines

Recommendation engines are what power the product recommendations. The key role of the engine is to analyze large amounts of data and select the best products to show to each unique user. Since a recommendation engine can increase the likelihood of generating results, it’s important to choose the right system based on your requirements.

Collaborative filtering systems

Collaborative filtering systems look at user preferences and behavior. This data is used to recommend items to other users who share similar preferences and behavior. What makes collaborative filtering unique is that it doesn't consider the actual content or products that users engage with when choosing what to recommend, it’s all based on user behavior instead.

Collaborative filtering systems can be either memory-based or model-based. Memory-based systems take clusters of users and examine their interactions to predict the preferences of similar users. Whereas model-based systems use machine learning and data mining to make these predictions. 

Content-based filtering systems

Content-based filtering systems personalize recommendations based on product features and attributes, making it an effective way to show new and alternative products. Examples of this could include “We think you’ll love” or “If you like that, you’ll like this”.  

The filtering system groups similar items by attributes such as genre, type, color, season and style and the algorithm is able to consider both customer preferences and item descriptions to generate the recommendations. For example, protein bars and shakers may be recommended alongside protein powder since all of these products share commonalities. 

Hybrid recommendation systems

Hybrid recommendation systems combine both memory and content collaborative filtering, benefitting from both types of systems and offering more comprehensive suggestions. Hybrid systems can be used in ecommerce by taking into account user interests with collaborative filtering, as well as product features and attributes, which is content-based filtering.

Optimize your product feeds 10x faster with AI
 

Personalized product recommendations examples

Here are some examples of personalized product recommendations that use AI-powered technology to enhance the shopping experience.  

1. Recently viewed items

Luxury retailer End Clothing features product recommendations by highlighting recently viewed items, which are displayed on product pages, category pages and even at the beginning of checkout. 

Putting the spotlight on items that a user has already engaged with acts as a reminder and encourages them to purchase the items.

recently viewed product recommendation

(Screenshot: endclothing.com)

2. Frequently bought together

Amazon personalized product recommendations feature bundles or items that are frequently purchased together, which is another great example of using a recommendation engine. The products displayed are tailored to users based on what they are interested in

This helps to speed up the time it takes to buy multiple items, it can highlight added benefits or potential savings if the bundles are discounted, and it can increase the size of an order. All of these things are beneficial for customer experience, AOV and conversion rate. 

frequently bought together example

(Screenshot: amazon.co.uk)

3. We think you’ll love

Another example where machine learning is used to capture attention is from online retailer boohoo.com, who feature products that individuals will be interested in. Using AI, the recommendation engine tailors products based on what users engage with on the website, or by using data on their interests and behavior. 

For example, if a user views sports sweaters, the engine can recommend top selling sweaters or outfits that feature sweaters. This level of relevance enhances customer experience and therefore engagement. Boohoo does this with a selection of products under the heading “We think you’ll love”.

we think you'll love example

(Screenshot: boohoo.com)

5. You might also like 

Multi-channel retailer in the building and construction sector, ToolStation, also highlights relevant products based on what a user engages with on the website. These appear towards the bottom of product pages under the heading “You might also like”. 

It’s an effective way for Tool Station to put emphasis on top selling products and highly reviewed products, something that adds social proof to the recommendations.

 

(Screenshot: toolstation.com)

6. A combination of recommendations 

Some retailers use more than one type of product recommendation on the same page, to further increase the chances of making a sale. 

In this example, fashion retailer Uniqlo showcases a selection of products based on the type of products the user shows interest in (sweaters), along with another recommendation that displays recently viewed items.

using a combination of product recommendations

(Screenshot: uniqlo.com)

7. Complete your routine / complete the look 

Another great example of personalized product recommendations is from beauty and skincare brand Kiehl’s. On a product page for fash wash, Kiehl’s recommends products to ‘complete your routine’. This brings visibility to relevant products that are likely of interest, improves customer experience and can help increase AOV with cross-selling. 

(Screenshot: kiehls.co.uk)

8. Email recommendations 

Using personalized product recommendations in email is another great example of how they can be used to boost conversion rate. One key way is to highlight new or alternative products to existing customers or newsletter subscribers based on their purchase history and interests. 

This could be on an ongoing basis, such as a weekly email newsletter, or at key moments by using data to determine when your audience are most likely to purchase again. 

product recommendations vis email

(Source)

These examples demonstrate the versatility of personalized recommendations, and how each ecommerce business can use them to their own advantage based on the buying habits of their customers. 

Optimize your product feeds 10x faster with AI

How to create effective product recommendations using AI

Create personalized product recommendations that use AI to generate results in five steps: 

1. Identify audience and behavior 

In order to accurately target recommendations and personalize shopping experiences, you must first identify audiences and their shopping behavior. This involves segmenting customers based on similarities between them or by similarities between product attributes. 

At a very broad level this could include grouping customers together based on demographics such as gender and age group, or by geo-location. A recommendation engine will then be able to show relevant products to these customer segments based on what they will most likely be interested in. The more granular you go, the greater the personalization. 

Product segmentation by grouping products that share characteristics or that complement one another is also an effective approach. For example, a pet store could group products together by animal so that dog owners are targeted with dog related products, and not cat nor fish products.

Or a recommendation engine could highlight dog accessories, treats or vitamins when a user is browsing dog food, if the engine predicts the customer is also interested in these types of products. 

your pet might also like example

(Screenshot: zooplus.co.uk)

 

Once audiences have been defined, it’s possible to form a strategy around how best to personalize product recommendations for your customers. 

2. Product recommendation placements 

The next step is to consider product recommendation placements, which refers to where products will be shown on your website, app or email. While there’s no right or wrong method, it is important to think about the customer journey and how it can be enhanced or improved through personalization.

Common places where personalized product recommendations can feature include: 

  • Category pages
  • Product pages
  • Shopping cart
  • Checkout process
  • 404 pages
  • Abandoned cart email
  • Weekly newsletter
  • App push notifications

Having full control over your channels means you can get as creative as you like, however, the art of product placements is to ensure they are clearly visible but not invasive. Including them on the homepage or bombarding consumers with recommendations across the entire site may devalue them, cause users to ignore them or they may even become irritating.

The best practice is to think about recommendation placements while taking into account the entire customer journey. 

3. Number of products that get recommended

When deciding how many products to include in the recommendation sections, consider factors such as industry, inventory size, and purchasing behavior. The type of product and the stage a consumer is in during the buyer journey also play a crucial role in deciding how many products to feature. 

For example, it may be best to showcase a wide selection of products early in the journey to give greater visibility to a range of different products. Alternatively, shopping cart or email recommendations that are nearer the end of the journey could benefit from featuring one or two highly targeted items to encourage final purchases. 

Tailoring the number of recommendations to these factors can enhance their effectiveness, which makes it an important part of the strategy. 

 

Optimize your product feeds in minutes with AI

 

4. A/B testing 

The next step in creating effective product recommendations using AI is A/B testing the strategies that are implemented. Performance will differ from one brand to the next so it’s important to test and learn what works best for you. 

Regularly A/B test everything in steps 1 to 3, such as how each of the customer segments responds to product recommendations and how different groups or clusters of products work together. 

Test the performance of different product recommendation placements to determine those that are the most effective for generating sales. Similarly, experiment with the number of products that are being recommended as well.

The aim of personalized product recommendations in ecommerce is to mimic the skill of a top salesperson and this can only be achieved by nurturing and developing the strategy. So if you have multiple ideas, test them all to determine the best approach to move forward with. 

5. Measure success with key metrics

The final step is to measure the success of your personalized product recommendations. It follows on from A/B testing different strategies and the best approach is to use a mix of metrics to understand the campaign’s success. Here’s what to focus on:

  • Product page views

Monitor the page views of products that are included as part of personalized recommendations. An increase in clicks to product pages could indicate an interested audience. 

  • Time on site

The aim of product recommendations is to improve user experience and by doing this, there should be an increase in the average time a user spends on the site.

  • Bounce rate

As with time on site, if bounce rate decreases following the launch of a product recommendation campaign, this could be an indication that it’s improved user engagement. 

  • Email CTR

Analyze CTR from emails and compare this with previous email CTR performance. Recommendations should make emails more engaging and result in a higher CTR. 

  • Increased average order value (AOV)

An increase in AOV following personalized product recommendations is another way to measure success. If AOV increases then it could mean the campaign is successful in upselling and cross-selling.

  • Uplift in sales

It’s hard to know if a sale would have taken place with or without the recommendations, however, a general uplift in sales is another good measure of success.

Optimize your product feeds 10x faster with AI

Key takeaway

Personalized product recommendations are a powerful way to enhance customer experience and drive sales. Recommendation engines get to know a customer's tastes and preferences inside out, helping businesses to achieve personalization that encourages brand loyalty and makes online shopping that much more enjoyable. 

Choosing the right recommendation engine, ongoing optimization, and robust A/B testing are crucial steps for achieving successful results. Personalized recommendations not only boost revenue but also demonstrate to customers that their unique preferences are understood and valued, making it a great addition to your ecommerce strategy.   

About DataFeedWatch

DataFeedWatch by Cart.com is the top-rated global product feed management platform that enables eCommerce brands, retailers, and digital agencies to drive multichannel growth. Merchants on both custom solutions and popular shopping carts, like Shopify, WooCommerce, and Magento (among others), can choose from 2000+ integrated shopping channels, affiliate networks, and marketplaces in 60 countries (such as Google, Facebook, Criteo, Amazon, and more). Major global brands like adidas, Decathlon, and KENZO have used DataFeedWatch to improve product performance across channels and expand to new markets. Sign up for a free trial today and receive guided onboarding to get started.

Made with   by 

Write for Us