Optimising content recommendations to improve engagement with AI

Academic publisher

We launched a recommendation system powered by AI to increase engagement and improve the search experience.  

500%

Increase in click-through rate

190%

Increase in downloads

Background

A global academic publisher wanted to increase engagement by suggesting relevant content to researchers using its platform. However, with a database of millions of articles, manual discovery isn’t easy, and previous recommendations resulted in low click-through rates. 

Solution

We implemented a live, AI-powered recommendation system to serve articles similar to the source article or content popular within the topic. Deep content representations, created using all the data about the article, are used to identify the best recommendations. A reinforcement learning model learns what recommendations are the most relevant, and continues to optimise results based on click-throughs and downloads.

Impact

The recommendation engine provides a more engaging user experience, resulting in a 500% increase in click-through rates and a 190% increase in downloads. Recommendations can be sent by email to re-engage users and the system scales as the content library grows.

background placeholder
Elevated

“I’m really proud of this work, which delivered excellent results for our customer and helps to support academic research worldwide.” 

Ben Bedford, Head of Delivery & Product Management

Faculty

Experience.