Rumors swirled recently that the iPhone was coming to Verizon. Social media outlets quickly garbled up the leaks, and opinions flared across the Web about the news. For companies and marketers, that information is incredibly valuable, but seriously hard to track. How can opinion on Facebook and Twitter ever be accurately measured? Enter Crimson Hexagon, a social media monitoring and analysis firm that has helped Microsoft, HP, and CNN gauge public perception.
After the Verizon iPhone news hit, Crimson Hexagon analyzed online conversations about the subject using its unique "statistical human-assisted approach." Developed at Harvard's Institute for Quantitative Social Science, the technology originally began as the "hyper-accurate estimation, classification, and quantification of unstructured data," says company CEO Scott Centurino, which, in non-science speak, just means the measurement of public opinion from unsolicited data. The company uses sophisticated algorithms and dynamic dashboards to dig through and track massive amounts of information in real-time.
Their approach is simple. Companies first design a question that needs answering. In this instance, Verizon might have wondered, "How did consumers react to the iPhone rumors?" After selecting a content base to analyze (blogs, tweets, etc.), Crimson sets up a series of categories for possible public reactions, and then begins to feed each category sample posts. So, for its Verizon analysis, categories included: (A) I'm sticking with AT&T, (B) I'm switching to Verizon for the iPhone, (C) I'm considering switching to Verizon for the iPhone, and (D) The news was just another "crazy rumor."
Crimson then begins manually building a sample pool of content, taking loads of tweets or Facebook posts, classifying them into each category, and coaching its analytics tools to replicate the pattern. Over the days following the iPhone rumors, for example, Crimson Hexagon analyzed more than 27,000 conversations on Twitter, finding that only 12% said they'd stay with AT&T, and that close to 60% said they would or were considering a switch to Verizon for Apple's popular device.
"The user wants the equivalent of having read every post online, and carefully and accurately putting them in the appropriate category," explains Centurino. "We create the leverage to allow humans to do that. We help users scale their judgment and provide a context of analysis."
The information is broken down in a wide range of dashboards, and can be viewed from surface positive, neutral, or negative opinion, or zeroed in on for more specific reactions.
Unlike Crimson Hexagon, most social media monitoring companies reply on two common solutions: keywords and semantics. Both, says Centurino, only offer non-specific positive or negative portraits of public opinion and are severely limited. Keywords depend on an expansive library of definitions ("It's like Cirque du Soleil contortionism," he says), and a semantic approach (that is, the analysis of phrases and expressions) depends on a language model that recognizes sarcasm, snark, abbreviations, and an endless amount of Web slang.
Centurino says the analysis runs as high as 97% accuracy, and that Crimson Hexagon often runs validation studies to ensure so. CNN is using the technology to gauge public opinion for the mid-term elections, and Microsoft's Bing team now uses it to monitor reaction to product roll-outs.
"There's a real value in tapping into these millions or billions of conversations online, where people are talking openly and honestly about what's important to them," says Centurino. "Our real kick comes from being able to understand the underlying themes from those conversations, so you can actually act on them. If all I know is that negative things are being said, what can I possibly do with that information?"