News

Canvs study ties emotional reactions in TV-related tweets to viewership changes

canvs

Can TV-related tweets be used to predict viewership increases or decreases? Like the existence of Sasquatch, this question has continued to plague man.

While the jury is still out, Canvs today released findings from the largest TV viewership study using Twitter data. The study looked at the emotional responses (Canvs’s forte) of nearly six thousand TV episodes across genres and found that certain emotional responses featured in tweets within specific genres can be used to predict whether viewership of the show will increase or decrease for the next episode.

For example, “hate” emotions expressed in tweets during reality shows or dramas is the best predictor of whether viewership will increase next episode. For comedies, “love” and “beautiful” are more powerful indicators than “funny.” For every 1 per cent increase in “love” and “beautiful” reactions, there is a 0.1 per cent and .3 percent increase, respectively, in viewership the next episode.

“Emotional analytics advance our understanding of how audiences feel about programming, and as such, we should be able to use emotions to predict program viewership,” said Sam Hui, PhD and Chief Scientist at Canvs. “This analysis conclusively demonstrates that Canvs’ emotional metrics can predict the likelihood that the viewership of the next episode of a program will go up or down.”

Canvs has productized this research in the form of Canvs Viewership Probability (CVP), which uses Nielsen data found on TV by the Numbers. CVP scores for shows are now available to Canvs clients via the dashboard and API.

Get the Newsletter

Keep up to date with the latest news and insights.

Subscribe

Adam Flomenbaum

Adam Flomenbaum is a global leading expert on social television, over the top platforms, TV Everywhere, and content marketing. Flomenbaum has written for Dime Magazine and Brooklyn’s Finest (ESPN TrueHoop). In January 2014, he began contributing to LostRemote, a leading social TV blog, and was named editor in October 2014.

All by Adam