FORT LAUDERDALE, FL: Media brands that produce political polls should apply the same level of sophistication to this type of analysis as when making strategic decisions about their own businesses, a leading executive from CNN has argued.

Robin Garfield, SVP/Research & Scheduling at CNN, discussed this topic at the Media Insights & Engagement Conference 2017, an event held by KNect365.

In bringing some clarity to the pitfalls of drawing definitive conclusions from surveys, as traditional pollsters frequently do, she referenced CNN's own brand and audience studies.

"I look at audience behaviour; I look at consumer insights; I work very closely with the public opinion polling group … And I think it's really important to use all these resources to get the full picture of what's happening," said Garfield. (For more details, read Warc's free-to-access report: CNN tackles the problems with polls.)

"Would I ever tell Jeff Zucker, who runs CNN, to make a major decision based on our tracking poll? Of course I wouldn't do that.

"We need to use a variety of sources. We need to use all sorts of information. Our expectations are too great that a tracking poll would be able to predict the president."

As evidence that TV ratings are easier to foretell than outcomes at the ballot box, Garfield cited CNN's predictions for its average audience among 25–54-year-olds across last year's presidential election day as a whole.

Its pre-event forecast pegged this figure at 2,535,000 – and the actual total was 2,541,000. "We did amazingly [well] projecting the election [viewing]," she said.

"So why was that so easy? We're using ratings to predict other ratings. That's not nearly the task we have when we're trying to predict elections … We're able to estimate, and predict well, ratings.

"But we're using the same metric to project the [election] results. These are completely different things. Also, polls are a moment in time. And we're not doing polling on election day."

Data sourced from Warc