Users often do not watch an online video content in its entirety, and abort the video before it is completed. This is captured by the notion of audience retention rate, which indicates the portion of a video users watch on average. A decentralized coded caching scheme, called partial coded caching (PCC), is proposed here to take into account both the popularity, and the audience retention rate of the video files in a database. The achievable average delivery rate of PCC is characterised over all possible demand combinations. Two different cache allocation schemes, called the optimal cache allocation (OCA) and the popularity based cache allocation (PCA), are proposed to allocate cache capacities among the different chunks of video files. Numerical results validate that the proposed coded caching scheme, either with the OCA or the PCA, outperforms conventional uncoded caching, as well as the state-of-the-art coded caching schemes that consider only file popularities.