The Limits of Social Media Analytics and the Need for Human Analysis

Even with the right tools in place, social media veterans agree, there's significant value in having people -- not just software -- monitor social media comments. Frank Eliason, SVP of social media at Citi, for example, has appointed a person to monitor the thousands of mentions of Citi online per day because of nuances in meaning, such as sarcasm, that the bank's solution might not recognize.

"Tools that measure sentiment don't usually do a good job at measuring sarcasm, and there's a lot of sarcasm in social media," Eliason says. "Boy, do I love my bank," could be interpreted by software as positive, but a human looking at a person's history of complaints would realize they mean the opposite, he explains.

"My favorite way to [measure sentiment] today is still manually ranking comments," Eliason adds. "That's how you get your best and most accurate measurements. You can also filter through the noise the social web brings."

First Tennessee chief marketing officer Dan Marks also believes it's necessary to have a person monitor social media comments, though not necessarily full time. "The nature of monitoring is much less precise than, say, querying a database for accounts," he says. "It involves some level of trial and error and sophisticated understanding."

For instance, software searches for "1st Tennessee" bring up references to a local sports team and a volunteer regiment, Marks notes; a person would immediately know these are not relevant.