Netflix Mobile Personalization Teardown: Row-Level A/B and Micro-Genres

Design · 6 min read

Netflix increased personalization granularity by experimenting at the row level and dynamically generating micro-genres based on viewing clusters. Rows can now be reweighted per user session, which allows the app to surface niche categories like 'quiet documentary flicks' for matching tastes.

Card creative optimization uses localized thumbnails and short autoplay clips that are served by a creative ranking model trained on skip and click-through signals. These microtests showed that tailored creatives can lift CTR by up to 12 percent for marginal content, but running too many variants increased cache churn on CDN edges.

Editorial curation remains important: Netflix combines algorithmic rows with promoted editorial banners that are manually curated for launches. The hybrid model preserves brand control and allows editors to seed discovery that algorithms can amplify.

The product lesson is to treat UI rows as experimental canvases and invest in efficient creative cache strategies. Personalization at the micro-genre level delivers stronger personalization but requires disciplined infrastructure planning to prevent CDN and client cache bloat.