Challenge
The platform had a broad film and series catalog available on web, mobile, and smart TV, and a loyal user base, but viewer engagement lagged behind larger competitors. The main challenges included:
Low engagement and watch time: Users spent less time on the platform than on competing services and often stopped after only a few titles.
Outdated, rule-based recommendations: "Trending", "new releases", and simple history-based picks couldn't match user tastes or drive deeper discovery.
Limited segmentation: The client relied on just 5–10 predefined segments that didn't reflect real behavior patterns across diverse audiences.
Fragmented data: Viewing history, ratings, and metadata were stored in separate systems, limiting the ability to build complete user profiles.
No internal AI/ML capability: The client lacked the expertise to modernize their recommendation engine and needed a partner to design and implement an ML-based solution without disrupting the existing product.
Solution
ZONE3000 designed and implemented a new personalization layer to improve content discovery and viewer engagement across the platform:
Unified data foundation
Consolidated viewing history, metadata, ratings, and user behavior into PostgreSQL to enable consistent profiles and model training.
ML-based recommendations
Developed Python-powered models to generate personalized "watch next" suggestions based on viewing patterns and content similarity.
Advanced segmentation
Expanded segmentation from 5–10 static groups to dynamic micro-segments tailored to user behavior and preferences.
Integration with engagement tools
Connected recommendation outputs to SAS Campaign to automate targeted re-engagement workflows.
Front-end personalization
Embedded personalized content sections into the React-based interface across web, mobile, and smart TV.
Data pipelines and monitoring
Built ETL workflows for regular data updates and feedback loops, ensuring stable model performance over time.
Technology used
AI/ML framework: Python to build and train recommendation models and analyze viewer behavior patterns.
Data & storage: PostgreSQL for consolidating viewing history, metadata, and user activity into a unified database.
ETL & automation: Python-based pipelines to ingest, clean, and update data regularly from multiple sources.
Engagement & campaign management: SAS Campaign to deliver targeted viewer notifications and re-engagement workflows.
Frontend & integration: React for embedding personalized content sections and recommendation widgets across web and mobile interfaces.
Result
The new recommendation engine and data workflows significantly improved viewer engagement and the overall content discovery experience:
More engagement with recommended content
Views from the Recommended section increased by 14%, showing higher relevance of suggestions.
Broader content discovery
Users viewed 6% more titles across the catalog instead of repeating the same top content.
Higher user satisfaction
Recommended videos received 16% higher average ratings compared to other sections.
Longer viewing sessions
The average time spent on the platform grew by 12%, reflecting more relevant content journeys.
Higher first-week retention
The number of users returning within a week of their first visit increased by 22%.
This case demonstrates how ZONE3000 improved personalization, strengthened user engagement, and helped a European VOD platform deliver a more relevant and predictable viewer experience.