Embodiments of the present invention provide an automated scheme for identifying high/low value content. Playback data from users in a social network may be periodically submitted by users of the social network. The playback data indicates segments of content, such as audio or video that the user has skipped over or tagged as low value. The playback data may also indicate segments of content that the user has repeated or tagged as high value. The playback data is then analyzed in aggregate and various clips are identified. In addition, the playback data may be compiled and organized among the users for future use. The playback data may be used to indicate segments of high/low interest to peers in the social network or to arbitrary users.