Make way for ai-powered content recommendation

Home » Case Studies » BFSI » Make way for ai-powered content recommendation


One of India’s leading media conglomerates, owning some of the country’s largest digital publishing plat- forms for news, lifestyle, and entertainment content, wanted to improve their audience engagement across multiple online platforms. A positive audience experi- ence, in terms of people being able to discover and con- sume relevant content, was an important driver of audi- ence engagement, leading to increased average adver- tising revenue per visitor.

The client’s digital publishing platforms hosted a variety of content including news, editorials, critics’ reviews, picture and video galleries. Additionally, the client had launched a web and mobile app for OTT media discov- ery, where users could view content suggestions across popular streaming platforms.


A content recommendation system was developed for five digital properties, to offer personalized content suggestions to website/app/platform users. The recom- mender system algorithm was trained using a combina- tion of unsupervised machine learning and reinforce- ment learning. It provided content suggestions upon processing live, click-stream data.


The following sources of data were used for model training:

  • Content Management System: article characteristics such as sections, sub-section, authors, header, estimated time to read an article, etc.)
  • Click through/ Click Stream: The I.P. from which an end-user has clicked at the new article, user’s location, reading habits, etc.
  • News Content: Tone and sentiment (derived through NLP), keywords extraction, etc.


Hybrid training approach was used for the recommender system:

  • Content-based: Keywords-based vector space models were developed to recommend articles similar to the one being currently read
  • User-based Collaborative Filtering (UBCF): An unsupervised machine learning method that recommends content by analyzing the preferences of similar users


Improvement in Audience Engagement Across Platforms

Key audience engagement metrics such as the click-through rate (CTR), page views, and bounce rate showed improvement for all the platforms where the recommender system was adopted. The ob-served improvement in CTR ranged from 6 to 8%

Download Case Study