Today’s Internet content providers primarily use Content Delivery Networks (CDNs) to deliver content to end-users with the aim to enhance their Web access experience. We adopt a measurement-based approach to understanding why, how, and how much CP-level intelligences can help content delivery. We first present a measurement study of the CDN built by Tencent, a largest content provider based in China. In contrast, knowing the exact context under which contents are consumed, CPCDN is able to exploit context intelligence capturing how contents are dynamically processed and synthesized in different contexts, to optimize the user experience and service quality along content delivery. In this paper, to accompany the rapid move to CPCDN, we apply a measurement-based approach to understand why, how and how much CP-level intelligence can help content delivery. Based on the crowd patterns, we proactively replicate contents according to the interest groups so that users can fetch the contents they are interested in from nearby CPCDN servers. Leveraging the powerful user intelligence capturing how contents are dynamically shared and distributed among social-networked users, CPCDN is able to improve both the content delivery efficiency and user experience to the next level.