Assembling streamed content for on-demand presentation is described. In embodiments, a media asset can be requested from a content distributor, and cacheable segments of the media asset are streamed to the requesting client device. The cacheable segments are received in an initial quality display format for uninterrupted playback of the media asset at the client device while the cacheable segments are being received. An asset matrix that correlates a bit rate and a time position for each cacheable segment in the media asset is also received. The cacheable segments of the media asset can be assembled according to the asset matrix to generate an assembled asset data file of the media asset. The assembled asset data file can then be back-filled with selected ones of the cacheable segments that are replaced with a highest quality display format for on-demand presentation of the media asset when selected for viewing.