• Canvs Finds That Viewers’ Feelings of Hate Are Best to Drive TV Ratings

    There are more scripted TV shows being made than ever, by one recent count over 400 in 2015, up 10x in the past 10 years. In this sea of choices, how can networks attract viewers and keep them watching? Well, a new study by Canvs suggests that eliciting feelings of hate toward certain characters is the most likely predictor of increased viewership for the show’s subsequent episode.

    Canvs said the study is the largest one ever analyzing the correlation between viewership and Twitter data. Canvs is a startup that uses language analytics to detect a range emotions contained in social media, which it then sorts into 56 different categories, such as “hate,” “excited,” “love,” “happy,” etc.). In the study, Canvs looked at tweets related to 5,709 episodes of 432 comedy, reality and drama shows that aired between January, 2014 and June, 2015 across broadcast, ad-supported cable and premium cable networks.

    According to Jared Feldman, Canvs’s CEO and co-founder, “hate” emerged as the most important predictor of following week viewership for reality and drama series. For every 1% increase in “hate” sentiment, Canvs found a .7% increase in adult 18-49 viewership of the subsequent episode as measured by Nielsen and reported by TVbytheNumbers.com. For reality shows that was higher than “crazy” (at .3% and “love” at .2%). For drama that was higher than “funny” and “love” (each at .3%) and “crazy” at .2%.

    For Comedy, surprisingly, “funny” didn’t register as the most important predictor of subsequent viewership. Rather, “beautiful,” which measures the attractiveness of characters, was the most predictive - a 1% increase in “beautiful” led to a .3% increase in viewership of the subsequent episode. “Love,” which followed, registered a .1% increase in viewership.

    To capitalize on the findings, Canvs also introduced a new scoring system it has dubbed Canvs Viewership Probability (“CVP”) which uses the same methodology of using Twitter data to predict how well an episode of a show will be in driving viewership of a following episode. Jared said 5 cable networks are currently using CVP.

    The Canvs data is so useful to networks because it provides scene-level feedback on what’s resonating with audiences and what’s not, at least as measured by social reactions. Though those tweeting during live airings are just a sliver of a show’s total audience, their reactions provide fresh insights about the show’s resonance.

    Jared said networks are already using Canvs data to create ancillary content for social distribution, determine the content of tune-in ads, and help develop plot points. With hundreds of new TV shows now being made and billions of dollars being invested, it’s easy to see why additional insights from Canvs on why a show succeeds or doesn’t are valuable.

    More information about the study is here.