Feature Guide

What is a Product
Recommendation Engine?

Your guide to Attraqt’s Product Recommendation capabilities

Download the guide →

Product recommendation explained

In the context of ecommerce, a product recommendation means showing users items they are likely to love and promoting relevant upselling and cross-selling opportunities.

This involves collating and analysing a whole host of customer data points like browser behaviour, past purchases, product preferences and demographic information to make intelligent, relevant predictions about what customers want.

In the omnichannel space, items suggested by product recommendation algorithms may be given prominence on websites, apps, social feeds, direct sales emails and more.


Product recommendation works

7 %

Revenue generated by Attraqt’s Gift Finder for French retailer Nature & Découvertes

9 %

Of turnover now generated by Attraqt’s product recommendation engine for Nature & Découvertes

40 %

Increase in CTR from product recommendations for French fashion retailer La Redoute

10-powerful-algorithms-stage-2-1 2

The product recommendation process

Nature & Découvertes’ Gift Finder lets users select from broad criteria of choices to narrow their search for the perfect gift from the site’s +30K inventory.

Users start by describing the type of gift they are looking for and can use playful language to talk about its recipient. The Gift Finder then uses a decision tree workflow to guide shoppers and makes suggestions in a way that feels remarkably similar to talking to a store assistant.

The result is a personalised and quirky shopping experience that yields consistently relevant yet unique gift suggestions.

However, the true power of product recommendation engines is in automated suggestions displayed in search results, product listing pages (PLPs), checkout pages and other digital retail spaces.

These could be accessories for in-basket items, upselling and cross-selling opportunities or suggestions for entirely new products.

Such targeted product recommendations can only be carried out by data collection systems and profiling algorithms working in unison to collate and analyse customer information to display products from the site’s inventory in real-time.

Approaches and algorithms


Content-Based Filtering

Perhaps the most basic product recommendation method, content-based filtering matches product keywords against the user profile. For example, a content-based product recommendation engine could show Kurt Geiger handbags to a shopper who has previously bought shoes from the same designer.

Content-based filtering pros
  • Doesn’t require large user data sets and is relatively simple to implement and scale from a ‘cold start’ where little or no user information is initially available
  • Caters to shoppers’ individual niche interests
Content-based filtering cons
  • Relies on some human knowledge and manual input to describe product feature sets. Content-based filtering can only be as accurate as the available information
  • Only makes product recommendations based on users’ existing interests and doesn’t make predictions about other preferences

Collaborative filtering

Collaborative filtering recommends products by rating the similarity between user profiles (user-user filtering) or items based on ratings (item-item filtering). User-user filtering aggregates and compares user data to suggest items based on the buyer’s similarity to other profiles.

Item-to-item collaborative filtering uses a similar approach. But instead of the algorithm identifying similar profiles, it identifies related products by category, brand, style, colour, and more.

Collaborative filtering pros
  • Yields accurate recommendations without having to extract, analyse and ‘understand’ products and their contextual relationships
  • Captures consumers’ changing interests over time
  • May generate relevant but unexpected and seemingly unrelated product recommendations
Collaborative filtering cons
  • Requires accurate and extensive user datasets to get started
For example:

If User One buys items A, B, C and D, and User Two purchases items A, B and C, collaborative filtering predicts the second user may be interested in buying item D in the future.

testing-2x 3
10-powerful-algorithms-stage-2-1 2-2

Taking a hybrid approach

Attraqt’s product recommendation engine uses a combination of user-user and item-item collaborative filtering with content-based filtering. Users can use algorithm stacking and automated A-B testing to find and fine-tune the most effective approach.

Teams can implement hybrid strategies from a ‘cold start’ without extensive data to set up. Moreover, they have been shown to yield consistently more accurate recommendations than a purely collaborative or content-based approach.

Product recommendation challenges

Product recommendation engines increase AOV, drive revenue and encourage brand loyalty. However, setting up efficient recommendation processes does pose challenges.

Syntax and Synonyms

Products often have more than one name or listing to cater to users’ search  language – for example, handbag, cross-body bag or clutch.

Attraqt’s product recommendation engine uses AI and natural language processing to understand search terms without the need for merchandisers to set up synonymous listings. It understands synonyms, slang, stop-words, spelling mistakes and more.


As inventories, customer bases, sales channels and datasets expand, product recommendation systems can be challenging to scale.

Attraqt’s eCommerce personalisation engine provides several ‘cold start’ strategies that don’t rely on historical data to implement. Teams can integrate and get started with advanced omnichannel personalisation strategies from the get-go.

Data Sparsity and Cold Start

Product recommendation engines lack new-user and product data.

Attraqt’s out-of-the-box product recommendation engine features ‘cold start’ algorithms teams can use to generate relevant suggestions without extensive historical data. Ecommerce providers easily can set up and manage effective recommendation strategies from day one.

The power of personalisation


Enhance insight

As Attraqt’s product recommendation engine collates and analyses data, it identifies trends and customer segments that trained human eyes can’t see.

Boost revenue

Similar items, accessories, and cross-selling and up-selling opportunities generated by Attraqt’s product recommendation engine consistently achieve an impressive uptick in CTR and AOV for our ecommerce partners.


Achieve consistent omnichannel messaging

ECommerce brands need more than just repeat sales. Personalisation improves the customer experience, generating meaningful transactions and brand loyalty in a world where they are becoming ever more scarce.

Build brand loyalty

In today’s fiercely competitive ecommerce landscape, brand affinity is quick to shift. Product recommendations and other personalisation strategies build strong personal brand-buyer relationships that last.


Now see it in action

Want to find out what Attraqt’s product recommendation engine could do for your ecommerce brand? Ask us about seeing a demo today.