There's the stuff we say on social media. And then there's the stuff we really mean. Marketers are determined to figure out the latter.
For several years, companies have been listening to people's chatter on Twitter, Facebook and other outlets to gather data about their likes and dislikes. Now they're tapping new software that helps them act on the data they're culling.
This technology can go beyond figuring out the literal meaning of hastily thumb-typed messages, to understand what people are actually saying—as a human being would in a real conversation filled with cues like intonation and body language.
Using this process—known as sentiment analysis or sometimes conversation analysis—marketers can better identify people who might be interested in their product or figure out which customer complaints demand an immediate response. And they can act on that information quicker than ever—often at the very moment that a subject related to their product is trending heavily.
Listening Closely
The software that marketers use works on a couple of levels. First, it breaks down tweets and status updates to figure out the literal meaning of what's being said—a step called natural-language processing. But then the software has to figure out the emotions behind the statement. Did somebody write it in earnest, or was it snarky? Was the writer showing a particularly strong emotion about the statement—enthusiasm, anger, sadness?
That's where sentiment analysis comes in. Programmers might give their computers thousands of examples of sentences and assign a mood to each one. The computers digest those and learn to look for the same types of patterns on their own.
These tools are being used by marketers and analytic companies in any number of ways.
To understand how patients felt about certain cancer drugs, social-analytics company BehaviorMatrix LLC, of Blue Bell, Pa., data-mined millions of cancer blogs. Anytime one of the drugs was mentioned, BehaviorMatrix categorized the emotion expressed as one of fear, acceptance, grief, optimism or despair.
Analysis of chatter about Kelly Osbourne's and Hayden Panettiere's hairdos at the Golden Globes helped guide a L'oreal promo on Twitter. Getty Images
Chief Executive William Thompson says he discovered that right when a person is first diagnosed with cancer, they are the most optimistic. So, he says he advises pharmaceutical clients to target ads based on the emotion the person is experiencing in the moment.
Another recent campaign unfolded around the Golden Globe Awards. When Hayden Panettiere and Kelly Osbourne walked the red carpet, their slicked-back hairdos drew detailed remarks from thousands of women who tweeted as they watched.
Almost instantaneously, they found in their Twitter feeds instructions from L'Oréal Paris showing them how to capture various red-carpet looks at home—as well as promotions and deals for L'Oréal products.
That took a lot of planning by L'Oréal and Poptip, a real-time market-research company in New York.
A few days before the ceremony, Poptip conducted a general analysis of what conversations about hairstyling were taking place connected to Golden Globe hashtags and other key phrases. Then, as the red-carpet events began, Poptips' software looked for similar chatter and analyzed which conversations were earnest discussions from the appropriate demographic.
"We're not targeting to males talking about how hot an actress is," says Poptip founder Kelsey Falter. Poptip determined that the target audience was captivated by slicked-back 'dos—so L'Oréal sponsored tweets to land in those Twitter conversations.
"You used to have to 'spray and pray,' " says Rachel Weiss, L'Oréal U.S.A.'s vice president for digital innovations, content and new ventures, "but now we don't have to guess anymore."
Handling Complaints
Crisis management is benefiting from sentiment analysis, as well. For example, just before it introduced a new product package, a granola company posted a photo of it on its Facebook page. The company didn't expect much feedback, but ended up with thousands of complaints.
Poptip's software analyzed the responses to determine if consumers were just bristling at change or if they were having a real emotional reaction, and where it was coming from. Poptip found that customers did, in fact, have a specific complaint about the new design: There were lots of comments referring to "the window" and "I like to be able to see" the product.
The problem: The older packaging had a window through which consumers could see the granola, and they liked that. The company got the message and halted production of the new wrapper, Ms. Falter says.
Sentiment analysis has also been used in political efforts. Earlier this year, the nonpartisan, nonprofit group No Labels—which seeks to strip partisanship from politics—teamed up with Los Angeles real-time analytics and trend-intelligence company Bottlenose. The goal was to find places to inject meaningful messages in measured, bipartisan Twitter conversations—which in part means hunting for tweets that eschew anger, sarcasm and other emotions that indicate a hard-line view of politics.
During the State of the Union address, Bottlenose analyzed the keywords being used in tweets hashtagged #SOTU, among other things. Adam Blumenfeld, Bottlenose's director of client digital strategies, says he and a team from No Labels found a subset of tweeters expressing support for bipartisanship were also using the hashtag #TeamUSA.
Into those conversations, they tweeted, "Both Democrats and Republicans are a part of #TeamUSA. We need a national strategic agenda. That's #HowWeFixIt." And they included an image with No Labels' Web address. In about 120 seconds, that tweet was retweeted 32 times.
But amid all the slicing and dicing of words and meaning, those who study consumer sentiment say that age-old rules still apply. "Our research has shown that there is a value in being talked about, positive or negative. It's better to have both than to have neither," says Mekkin Bjarnadottir, marketing manager for Lexalytics, maker of software that powers platforms like Bottlenose's.
Or as an old-fashioned press agent might have put it: Just tweet my name right.