The Complete Guide to Personalized Product Recommendations

Have you ever tried on clothes in a store and had an associate bring over a top that she says will look great with the jeans you’re considering? Or you are about to check out at a kitchenware store when the clerk recommends a cleaning product that will help you keep the cutting board you are purchasing looking new?

These clerks were trained to help close the sale and increase the average order value by making product recommendations. The recommendations are a mix of personalized (the shirt that goes with the jeans) or standard (the cleaning product to oil the cutting board). Offering recommendations is a tactic as old as retail itself.  

What are Product Recommendations?

In the eCommerce world, product recommendation engines use machine learning to select items to be displayed in a specific marketing channel (emails) or across the organization (web, email, and social).

And product recommendations work. Our customers tell us they see, on average, the following uplift from using recommendations:

Do all product recommendation engines deliver personalized recommendations?

The short answer: No.

The non-personalized recommendations offer shoppers a general recommendation. They use collaborative filtering techniques, social proof, and business rules to offer recommendations. They use a variety of data and techniques. Here’s how each works:   

Crowdsourced recommendations

Shoppers who looked at those salad bowls looked at these tongs.

Social proof

Top-rated items are displayed in the category viewed.

Business rules

Marketers elect to show items from a category such as top-selling items, items on sale, or items in low supply.

How are personalized recommendations different?

A personalized recommendation engine queues up product recommendations that are unique for every viewer or subscriber.

The engine uses the customer’s buying and browsing history to create a profile that leads to recommendations specific to that customer. If a shopper is clicking on turtlenecks, the recommendations engine will show turtlenecks that haven’t been viewed yet. It might also show similar brands (to those viewed) or similar categories.

If a customer has recently purchased furniture in a specific style, they would see furniture similar to that when they shop online, or a promotional email would include recommendations that match their recent purchase.

Personalized recommendations can also include location-based triggers - such as pulling in a real-time weather forecast into emails and website content and queuing up products that make sense for the weather forecast in the customer’s location.

Can one recommendations engine work across email and the online store?

Some recommendation engines only work for one channel. That means more work for the marketer. Setting up separate programs for email and the eCommerce site is time-consuming, especially if A/B testing and sign-offs from stakeholders are required.

Having one recommendation engine that sits on top of the different channels and delivers the appropriate content to each channel is a more logical choice.  

An omnichannel approach is particularly important for personalized recommendations. If a shopper has been clicking on certain clothing items but abandons the browsing session, you might send a browse abandonment email showing other choices in that category. But not everyone opens the browse abandonment email. If the shopper returns to your site, you’ll want to make sure the product feed reflects what they were interested in earlier.  

How can I measure the success of product recommendations?

Unlike a cart abandonment email program, measuring the success of product recommendations isn’t quite as simple as tallying the conversions from the campaign.

Here are some of the metrics that should improve after implementing a product recommendations program:

Changes in website browsing time

A product feed on the home page should be engaging the customer more than a generic one. But it might not increase browsing time. Instead, browsing time can go down in relation to conversions. Shoppers are finding what they are looking for more quickly.  

Higher click-throughs on email campaigns

Promotional and triggered campaigns (cart and browse abandonment) should be more engaging with recommendations.

Higher click rates on your website

If the product feed is personalized that should increase click rates as customers view what you’ve suggested for them.

Higher open rates for emails

Category information can be pulled into subject lines enticing shoppers to open the emails where the more specific recommendations are offered.

Increased average order value

If your recommendations are designed to encourage customers to buy products that go with the ones they are browsing or carting, this metric should increase.

Sales uplift

Whether from the email program or the overall eCommerce sales, a sales uplift if the best metric.

How are brands using product recommendations?

Large brands almost all use some type of product recommendations as a part of their online retail programs. It’s companies like Amazon, after all, that helped shape the success of eCommerce by pioneering product recommendations.

But the technique isn’t exclusive to large brands. Plenty of small to mid-size online retailers have successfully deployed product recommendations to increase sales.

Recommendations across channels

MyOptique Group incorporates personalized product recommendations on its website and in triggered and campaign emails. It’s Glasses Direct division attributes 1.2% of online revenue to web recommendations.

Recommendations in abandonment email

Orlebar Brown, a London-based fashion retailer, incorporated product recommendations into its cart abandonment emails. The move contributed to a 6.59% sales uplift. It also added recommendations to in-parcel marketing information sent with its items which adds a delightful bespoke touch.  

Subject lines that engage

Wolseley, a leading supplier of building materials, has increased its average open rate from 28% to 45% using automated product recommendations fed into subject lines.

Personalizing the home page

Global surfing brand Rip Curl saw sales increase 1.8% after adding personalized recommendations to its home page. The success has encouraged the company to roll out personalized recommendations to its product pages.

Social proof + recommendations drive click-throughs

Pet product specialist VioVet incorporated its in-house star rating system into all of its recommendations (and cart and browse abandonment emails). This has resulted in a 19% increase in click-throughs.

What should I look for in a product recommendation solution?

The popularity of product recommendations as a marketing tool has led to a lot of vague marketing boasts. At its heart, a good product recommendations solution is backed by solid machine learning and is easy for the marketer to use.

Here are some other features to look for in a solution.

Robust collaborative filtering

There are many different options for non-personalized recommendations. Make sure your engine includes a wide variety including options like "Frequently bought with this" and "After viewing this, people buy".


Some product recommendation solutions incorporate machine learning, but it’s only used to deliver collaborative filtering recommendations. Personalized recommendations use customer browsing and buying behavior.

The option to blend recommendation types

Sometimes you want to give customers recommendations based on what other customers bought - but with a twist. When you have access to both personalized and crowdsourced recommendations, you can set up both to provide variety and maximize the conversion potential.

A cross-channel rules engine

Recommendations work best when they are woven into different channels seamlessly. Solutions that only work for email or the web limit that option.

Real-time recommendations

To use recommendations in cart abandonment emails (which should be sent out within an hour), you need an engine that can pull the information quickly.  

Ease of use

Recommendation programs that are difficult to launch or involve complex integration steps will make it more difficult for marketers to use effectively. You’ll also want the option of reusing templates.

A final word on product recommendations

Cart and browse abandonment strategies have been gaining ground, consistently showing returns for several years. Recommendations are the next logical step as machine learning drives more options and features every year.  

Yet, marketers are sometimes hesitant to add recommendations to their toolbox because they think setting up a recommendations program is too time-consuming.

Our clients tell us that once they take the plunge, those fears are allayed.   

A recommendations program can start simple - with recommendations going into an abandoned cart email, and gradually building out from there. It is a marketing tactic that not only increases revenue but also helps to establish that one-to-one relationship that customers say they want.

To find out more about Fresh Relevance, and our industry-leading Product Recommendations solution, click here.