A recommendation system predicts which titles a given viewer is most likely to watch and enjoy, then surfaces them - in rows, rankings, and artwork - so the catalog feels personal. For large libraries it is the primary way viewers find content; search and browsing handle only a fraction of discovery.
Recommenders blend techniques - collaborative filtering (people like you watched this), content-based filtering (similar to what you watched), and increasingly deep-learning models - fed by viewing history, context, and metadata. Because discovery drives engagement and engagement drives retention, the recommendation system is one of the highest-value pieces of an OTT platform: small improvements in relevance translate directly into watch time and lower churn.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
A recommendation system is how most viewers find titles in a large catalog, so its quality shows up directly in watch time. We wrote seven volumes covering the video stack; for recommendations start with Volume 6 on the ML basics behind product decisions, deployment topologies, and search over a video archive, then Volume 7 on serving and cost, with players in Volume 3. Written by Nikolay Sapunov, CEO at Fora Soft.
Viewers scrolling the catalog without finding anything to watch?
We will review your catalog, viewing data, and current ranking and send a recommendation system plan with the models and metrics to start with. 250+ video products shipped since 2005.
