A Coca-Cola Co. study finds online buzz has no measurable impact on short-term sales, but online display ads work about as well as TV, said a company executive in a presentation at the Advertising Research Foundation's Re:think 2013 conference in New York today.
It's a stunning admission for a company who's flagship brand has 61.5 million fans, more than any other brand on Facebook. But Eric Schmidt, senior manager-marketing strategy and insights at Coca-Cola, isn't giving up on buzz just yet. And he cautioned against reading too much into the research, noting that it covers only buzz, not sharing, video views or other aspects of social media.
But when Coca-Cola put buzz sentiment data into the same analytical framework it uses to evaluate other digital media, Mr. Schmidt said, "We didn't see any statistically significant relationship between our buzz and our short-term sales."
That was at a 95% confidence level, but even stepping back from that high standard, he said showed buzz affecting sales by only 0.01%.
"Is that the end of the story?" Mr. Schmidt asked. "I would say no. This is one study on a set of brands in a particular company within a certain segment of the consumer-packaged-goods industry. It is by no means a generalized result that applies to all industries. "
Now Mr. Schmidt said Coke is looking to refine how it measures buzz, for example by getting a better idea of how many people buzz actually reaches rather than just counting the raw publicly available comments from such sources as Facebook, Twitter, blogs and YouTube.
Coke research was far more favorable for digital display advertising, which it found on average to be 90% as effective as TV at generating sales on a per-impression basis, Mr. Schmidt said. Search was 50% as effective as TV – about the same as out-of-home – with radio coming in between TV and search and print scoring slightly more effective than TV.
One problem Coca-Cola has is determining whether buzz is actually positive or negative in the first place. In one 2010 study where Coke pulled out more than 1,000 social-media messages randomly and had human raters compare them to automated sentiment analysis by one vendor, there were widespread differences.
"When we say it's positive, the machine about 21% of the time says it's negative," he said. "That can cause some problems in our understanding" of how buzz impacts sales.
Machines have the most trouble judging sentiment in longer posts such as those in blogs or Facebook and do much better on Twitter, he said.
It's important to get digital reach and viewership data on par with TV and to accurately measure buzz in order to get more accurate return-on-investment analysis for all media, Mr. Schmidt said.
"Digital ROI has to be a financial measure that allows us a consistent measurement of business performance over time," he said, and also lets marketers make informed tradeoffs between media.