Statistical modelling of the size of the audience for televised football matches has utility for a number of stakeholders. For example, in the short-term it can inform the choice of broadcasters as to which matches to show and decisions of firms regarding how much to bid for advertising slots during the game or for displays on electronic pitchside boards at the stadium which will appear frequently on the television picture. For clubs and leagues at the elite end of the sport, for whom sale of television rights is their principal revenue source, it offers a means of understanding which match characteristics will draw the largest audience and this in turn should inform long-term decisions on how to structure their competitions. The paper argues that previous modelling in this area has typically depended on an inappropriate measure of audience size and unsatisfactory metrics to capture important characteristics of a match, such as the quality of the talent on show and the significance of the match for the championship or relegation. In contrast, the present paper draws on sport analytics to derive its measures of player talent and match significance. This proves to make a material difference to findings from modelling audience size for 790 English Premier League matches between 2013 and 2019. For example, it identifies elevated interest for matches potentially significant for end-of-season outcomes. This finding has implications when organisers consider the optimal structure of the competition and revenue sharing arrangements between member clubs of the league.