This is a study that asks: "How much value do Netflix-style ‘Because you watched…’ recommendations really create?" It measures what happens to user behavior and business outcomes when you turn personalized recommendations on vs. off.
Quantifies the true business impact of personalized recommendations (e.g., on viewing time, engagement, retention, and revenue), helping leaders decide how much to invest in recommendation systems and how to justify that spend with hard data.
For Netflix itself, the moat is a combination of proprietary user-behavior data at massive scale, mature experimentation culture (A/B testing infrastructure), and highly tuned recommendation algorithms that are difficult for new entrants to replicate quickly.
Classical-ML (Scikit/XGBoost)
Feature Store
High (Custom Models/Infra)
Model training and serving at Netflix scale (billions of interactions), plus experimentation cost when running large A/B tests across millions of users.
Early Majority
This work focuses on rigorous causal measurement of value (e.g., via experiments or quasi-experiments) rather than just proposing a new algorithm, making it directly useful for ROI and strategy decisions rather than purely for model accuracy benchmarks.
7 use cases in this application