Techniques for generating meaningful video-on-demand recommendations based on television viewing history data are described. Television viewing history data and video-on-demand purchase data are gathered from multiple client devices within a network. Television programs and videos-on-demand that are watched and purchased, respectively, using the same client device are associated with each other. Weights are assigned to the associations based on percentages or statistical analysis of the number of client devices through which particular videos-on-demand are purchased and television programs are watched. When a viewer requests video-on-demand recommendations, such recommendations are automatically generated based on comparisons between television viewing history data associated with the viewer and the television program/video-on-demand associations.