JioHotstar has published an engineering overview of its real-time ad request workflow, detailing how the platform coordinates distributed services to select and deliver personalized ads during video playback. The system uses a waterfall tiering approach combined with pacing algorithms (PID and SHALE) to select ads from thousands of candidates within a 100ms latency budget, even during high-concurrency live sports events. Key engineering challenges include managing inventory allocation, campaign pacing, failure handling, and cross-team coordination across ad decisioning, tracking, and analytics services. The architecture aligns with industry standards like OpenRTB while adding internal layers for personalization and content-aware targeting.
160 Impressions