Presidential Debates Provide Living Laboratory for Marketers Studying Sentiment Analysis

by   |   October 19, 2012 3:50 pm   |   0 Comments

Twitter debate data

Twitter correlated user activity during the Oct. 16 presidential debate to issues discussed by Mitt Romney and Barack Obama.

For those watching, the presidential debates have been full of drama, both on their TVs and on their digital devices. It is that second screen and the data generated by social media users, bloggers and forum participants that provides a living laboratory for sentiment analysis, and reinforces some fundamental lessons for marketers.

Related Stories

Improving the effectiveness of customer sentiment analysis.

       Read more»

Data research pays off at the London Olympics.

       Read more»

They know who you’re voting for: how big dta redefines political campaigns’ microtargeting.

       Read more»

While there are special characteristics to a live event focused on President Barack Obama and Mitt Romney and watched by more than 60 million people and generating more than 7 million tweets, the takeaways are the same whether it’s a debate, the Super Bowl or a targeted digital marketing campaign, experts say. Listen intently. Engage your audience when appropriate. Be ready to play defense. And scrutinize social media data for measurable insights while being aware of its limitations.

“The bottom line is it’s about engagement, tapping into your followers’ passion, and not assuming that the data is 100 percent accurate,” says Olga Spaic, manager of analytics for Metia, a digital marketing agency.

“Binders Full of Women”
This presidential campaign has had its share of Internet memes, phrases that propagate online riffs on a candidate’s remark (think of “we built it” hashtags and an “invisible Obama” Twitter account spreading during the Republican National Convention, or “Big Bird” on Twitter after the first debate). Social media users watching the second debate Oct. 16 created another one after Romney said “I brought us whole binders full of women” as he described his efforts as Massachusetts governor to recruit and hire women for jobs in his administration. (Media reports subsequently noted it was a group that submitted female candidates to Romney’s team.)

At this point in the debate, Twitter said its users were posting 104,704 tweets per minute. The remark inspired instant parody accounts on Twitter and Tumblr and several Facebook pages, creating a global online conversation about one phrase. It is this spontaneous, multi-directional maelstrom that generates both peril and opportunity.

“Here’s one phrase, and it was a dumb thing he said, and it was picked up not just by Jon Stewart and by political analysts, and it will have a much longer shelf life after it hits Tumblr and becomes a hashtag on Twitter,” says Paul Gillin, a social media marketing consultant and coauthor of Social Marketing to the Business Customer.

The peril in the binders episode is that the collection of Tumblrs, tweeters, Facebookers and bloggers create their own point of view, broadcast to multitudes, and it shows how marketers need to be wary that the voices of their brand are always subject to recording, reinterpretation, distortion and filtering.

“People will find hidden meaning and interpret what you say, and they have the ability to tell other people now, and it has the potential to start a small forest fire, which can occasionally become very big,” Gillin says. “One of the reasons you have to do this kind of monitoring is to find out when something happens outside your sphere of influence. When someone who works for you says something, you have to listen.”

Debate Data

Social media sentiment analysis notes that came out of the second presidential debate on Oct. 16:

  • Twitter: The peak tweets per minute tracked by Twitter reached 109,560 when users commented on former Gov. Mitt Romney’s slip when he said the name of an audience member asking a question on immigration. Other high points of sentiment came when Romney discussed tax rates and Obama criticized his opponent about his stance on China.
  • Facebook: The debate resonated most in the U.S. with men aged 35-44 and women aged 45-54. The District of Columbia, Maryland and Georgia Facebook users were the most engaged. The issue “economy/jobs/taxes” registered the highest on Facebook’s Talk Meter, 6.2 on a 10-point scale, and higher among men than women. “Binders full of women” registered a 5.
  • SAP: A social media analysis showed that nationally, Obama enjoyed “a much higher sentiment by the end of the debate compared to Romney.” In the U.S. as well as in swing states Florida, Ohio an Virginia, Romney had more users commenting on him—but it was not necessarily positive. In Ohio, an analysis showed Obama had a negative net sentiment throughout the day of the debate.
  • Ignite Social Media: This social media marketing company in Cary, N.C., found in its analysis that Obama had more positive social sentiment than Romney during the debate—but both were majority negative, with Romney showing 81.4 percent negative sentiment and Obama 68.7 percent.

But if a brand manager looks closely at what is going on, there could also be a subsidiary benefit to the maelstrom, says Terence Chesire, director of product management for social media solutions at SAP. Chesire says that an alert marketer can pick up cues about consumers’ perceptions of her brand by looking at the images mashed up on the Tumblr blog.  On Oct. 19, among the dozens of images were visual references to the “South Park” TV show, Meryl Streep in the film “The Devil Wears Prada,” and places where a consumer could buy binders such as OfficeMax, Staples and Walgreens.

“What you will see about a lot of these names [in the images] is that people are remixing and reusing a mix of images. There are incorporated brands in there. And if you are brand owner, this is where you roll with your punches. This meme is taking off, and your brand is taking off,” he says.

An Opportunity for Experimentation
Events like the debates that generate great public interest are also an opportunity for marketing managers to create social media marketing campaigns that connect the event to their brand, Spaic says. Noting that Seattle tilts liberal in politics, she cited the example of a local cupcake shop that ran a Facebook promotion coinciding with the debate for a flavored cupcake favored by Obama.

“If there is a connection to your business and you can use one of these trending hashtags or the event in general to get some buzz for yourself, then by all means do so,” Spaic says. “But tread carefully given the divisive nature of politics.”

And if a brand manager is sitting on the sidelines, the exercise of studying big events such as the debates provide a useful case study exercise, Chesire says. Looking at the sentiment analysis data from the debates can gives marketers a sense on how sentiments expressed in a multitude of channels trend positive or negative, what keywords and phrases generate interest.  It also can give marketers ideas for phrases that they can include in their own work.

“This tells you what phrases are resonating with people. If you have a little fun with it and don’t take yourself too seriously, you can use the language your customers are using to make your brand more relevant. Use the language and terms that your consumers are using already to make sure you are able to relate to them. Take a key phrase that came out [of an event like the debate]. How does that apply to your brand principles?” he says.

cupcake promotion tied to presidential debate

A Seattle bakery tied a Facebook marketing campaign to the debate.

SAP issued a report about the debates, citing its natural language processing analysis of more than 100 million data sources including social media, websites, blogs and online forums. Its findings: sentiment about Obama was more positive than that about Romney during and after the second debate. Romney tallied higher numbers of mentions, and showed a more passionate engagement across the country and in battle ground states like Florida and Ohio—but that was not necessarily positive.

Chesire says the natural language processing continues to improve as the technology matures—so that the tools can pick up nuance and tone in comments that are not meant as straight positive or negative. That is a well-known learning curve involved in sentiment analysis.  And while there is a high degree of certainty about location of social media users, and even gender, there is not much data about the age or other demographic aspects of users in SAP’s analysis. To get more specific, marketers typically correlate social media data with other demographic surveys and data sources.

That’s why the experts say brand managers need the human touch when looking at the real-time and post-event fallout from a big event. “Remember that machines can’t do it all—make sure the sentiment analysis  makes sense and adjust the numbers accordingly,” Spaic says.

Michael Goldberg is editor of Data Informed. Email him at

Tags: , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>