Predicting ratings for TV shows
A broadcaster wanted to learn if they could predict how future episodes and series of a programme would perform based upon historic performance.
What we did:
Ipsos Connect utilised social mentions about the programme and combined these with both TV ratings and some traditional survey metrics to create a statistical model that could help to predict the future performance of both episodes within a series and future series. We coded up over 425k Twitter mentions, about 118 episodes, covering seven series. Furthermore, there is the capability for the coded social mentions to be generated in near-real-time to an online dashboard, so the broadcaster and producers can see “live” updates of social mentions.
Social mentions strengthen the predictive model. Only via the inclusion of social mentions was the predictability raised to a level that could invoke confidence in predicting the performance of future shows, as the survey data alone could not offer the depth of feedback that social data offered.
This pilot has helped build awareness of the different ways social media data can be integrated with other data sources to provide incremental value within this broadcaster’s organisation.