Content delivery is a complex problem given the variety of services to be supported. Bandwidth availability and efficiency at high-end homes becomes key as high bitrate applications including 4K, 4K-HDR, and VR video become available. Performance is usually estimated in terms of latency and different types of latency measurement criteria could be considered. In order to achieve high performance across applications, we consider two aspects: modeling of viewership and efficient adaptive bitrate streaming delivery over IP. We delve into modeling viewer behavior using machine learning clustering techniques that use genre and demographics as features. The aim is to create effective content tranches to which bandwidth can be allocated more efficiently. We then examine the problem of adaptive bitrate streaming delivery of the content to the classes of viewers. In particular, we look at advanced encoding features at the video and transport level that can significantly improve user experience, while preserving video quality. We present the paper from an overall architectural perspective with deeper focus into the above described technical aspects.