Business Challenge
Mathrubhumi is a leading Indian print publisher and media house based out of Kerala with a strong presence across regional and national audiences. Known for its diverse content coverage, including news, lifestyle, and current affairs, the platform serves millions of users daily across web and mobile environments.
While the platform attracted substantial traffic, a large portion of users exited after consuming a single piece of content. Mathrubhumi aimed to enhance user engagement beyond the first article interaction.
Key challenges included: Limited content recirculation after initial article consumption, lower views per user and session depth, and lack of a structured mechanism to guide users toward relevant next content
Mathrubhumi faced the obstacles in driving audience growth and engagement due to limitations in their existing analytics setup. Firstly, the traditional set-up at their disposal provided only high-level, surface metrics such as page views and bounce rates, without the detailed insights necessary to understand reader behavior or content performance. This lack of actionable data left the editorial team with limited visibility into what stories truly resonated with their audience.
The goal was to create a seamless discovery experience that encourages users to continue exploring the platform.
The Approach
As a part of GNI Audience Growth Program, in collaboration with the Google News Initiative, Mediology Software and Tatvic, Mathrubhumi adopted a Measurement-First, Content Recommendation Engine-Led strategy to transform static content consumption into a continuous, personalized discovery experience.
The approach was built across three interconnected pillars:
Establishing a Reliable Measurement Foundation
The initiative began with strengthening the GA4 implementation to create a high-integrity data foundation. A detailed GA4 audit was conducted that helped Mathrubhumi identify and plug the key gaps in the implementation. This involved defining business-critical events like article engagements and user behaviour events to track user interactions across the content journey, Building advanced exploration reports to analyze acquisition, engagement, and audience behavior and Identifying key drop-off points, especially post article consumption. This enabled the team to move from fragmented insights to a clear, data-backed understanding of user behaviour, forming the basis for recommendation strategy.
Building & Deploying an Intelligent CRE
To address content drop-offs, Mathrubhumi implemented a Content Recommendation Engine (CRE) powered by a robust content intelligence layer: Established a data ingestion pipeline to extract and process article content, Applied text preprocessing and normalization to standardize content, Developed a content intelligence layer to identify underlying themes and relationships between articles, Continuously tested and refined the model to improve recommendation relevance before deployment.
This solution was built on a scalable Google Cloud Platform (GCP) environment, leveraging Gemini Enterprise Agent Platform for model development, along with Cloud Run, Cloud Storage (GCS), and BigQuery for efficient data processing and pipeline orchestration.
The CRE was then integrated within article detail pages, surfacing contextually relevant and trending content within the natural reading flow, effectively transforming exit points into continuous engagement opportunities.
Enabling Personalized Content Discovery
To further enhance engagement, the recommendation framework was extended with behavior-driven personalization: Introduced dynamic category recommendations aligned with user interests, Adapted content suggestions based on reading patterns and engagement signals, Shifted navigation from static browsing to a guided, intent-driven discovery journey.
This ensured that users were consistently directed toward the next best content, increasing both relevance and engagement depth.
