News Stay informed about the latest enterprise technology news and product updates.

Successful social media analytics requires commitment, experimentation

To get social media analytics right, companies should be prepared to experiment with multiple technologies and approaches.

Pradeep Kumar believes his investment in social media analytics will pay off … eventually. He just isn't sure how yet.

“It’s all in the experimental area now,” said Kumar, vice president and customer intelligence director at advertising firm DraftFCB, referring to the company’s efforts to mine and make sense of social media data like Facebook posts, tweets and blogs.

“We have only touched the tip of this iceberg,” Kumar said.

The field of social media analytics is a relatively new phenomenon and, with no playbook to follow, early adopters like DraftFCB are pretty much making up the rules as they go.

The potential benefits are an incentive to take the journey. By analyzing social media, companies could catch isolated customer service problems before they become systemic; identify and woo key “influencers,” bloggers and prolific Twitter and Facebook posters who can affect others’ buying behavior; and monitor how customers are taking (or not) to a new ad or marketing campaign.

But as companies start experimenting, some are realizing that the realities of social media analytics don’t always jibe with their preconceived notions. And the insights gleaned by the likes of Kumar could prove helpful to companies considering embarking on social media analytics.

Social media analytics requires multiple tools
While vendors like SAS Institute are releasing software designed specifically to perform social media analytics, Kumar has found that mining and analyzing social media data takes more than just one tool.

“Not any one vendor is offering that as a complete end-to-end solution,” Kumar said.

With dual headquarters in Chicago and New York, DraftFCB helps its clients devise, monitor and measure the impact of advertising campaigns. DraftFCB clients include GlaxoSmithKline and Kraft Foods, according to the company's website.

As part of its services, the company uses a combination of open source technology and vendor software to perform social media analytics, Kumar said.

Among them are tools to help identify which social media sites to monitor, others to collect and store the data, and still others to do the actual analysis.

The company uses technology from Radian6, for example, to identify which social media sites are worth monitoring in the first place, Kumar said. Radian6, along with software from MotiveQuest, also helps DraftFCB collect and do some analysis on the social media data.

For more sophisticated analysis, the company uses text analytics and sentiment analysis software from SPSS. In 2005, DraftFCB invested in software from SPSS, which was acquired by IBM last year, to analyze customer feedback from online forums and surveys. Turns out, text analytics and sentiment software is also the foundation of social media analytics.

“The tools give us the opportunity to analyze unstructured data in a lot more friendly, easier way,” Kumar said.

DraftFCB continues to experiment with different technologies to shore up its social media analytics capabilities, something most companies will likely need to do.

“Initially, you get a little confused,” Kumar said of navigating the different technologies available for social media analytics. Despite what vendors may say, “it’s not just one tool.”

Don’t forget the human touch
Another reality that confronted Kumar is that social media analytics requires more than just multiple tools and technologies. The truth is, sentiment analysis technology, while promising, is not terribly accurate at this point.

In fact, sentiment analysis tools currently on the market return “a hell of a lot” of false positives and false negatives, said Jim Kobielus, an analyst with Cambridge, Mass.-based Forrester Research.

The tools are not adept, for example, at picking up on sarcastic language or words with multiple meanings. Some words even have different meanings in different parts of the country, said Katie Paine, chief executive of KDPaine and Partners in Berlin, N.H.

Wicked is generally considered a negative term, for example, but in New England it’s sometimes used to express approval or praise, she said.

So in order for the tools to be most useful, organizations like DraftFCB have found it takes a human to direct, validate and sometimes correct the software’s analysis.

“There’s a lot of human interaction required to interpret what is coming out of these tools,” Kumar said.

While he wouldn’t discuss in detail any specific DraftFCB clients, he said just deciding which social media and networking sites to monitor is done largely by humans, not technology.

“We just create some basic lists,” Kumar said.

Humans are also needed to fine-tune the various rules that govern social media analytics and sentiment analysis tools. Here, too, Kumar said, DraftFCB works with its clients to draw up lists of positive and negative terms that are most likely to be used when describing products or services.

And when the software returns analysis, humans are needed to make sense of it and determine what actions to take base on it.

Execs at TelCentris, a cloud-based unified communications and VoIP provider, use technology from Viralheat to monitor mentions of the company on the Web. The tool gives Erik Bratt, TelCentris’ marketing chief, a view of the overall sentiment of tweets and other social media data related to the company.

To analyze individual tweets, however, Bratt and his team have to do that themselves. When he identifies a customer tweet complaining of a problem with TelCentris, Bratt replies with information on additional resources to help.

“I don’t think they’ve perfected [sentiment analysis] with machines yet,” Bratt said, referring to analytics vendors in general and not Viralheat specifically.

Until they do, said Forrester’s Kobielus, “you need a human being to eyeball it.”

There is, however, a lack of skilled business intelligence pros familiar with social media analytics tools, according to Kobielus. “You’re not going to find quite so many people off the street” who are experts in social media analytics technology, he said.

Social media analytics is a process
With the tools still evolving and business intelligence pros just starting to understand how they work, companies considering embarking on social media analytics projects should realize it’s an ongoing process, not a one-time event, Kumar said.

“It’s a learning curve for the analysts as well as the tools to make sure the right things are being considered,” Kumar said. ”It’s not something you could have the result right the first time.”

From tweaking which words to consider positive or negative to keeping up with the burgeoning blogosphere, companies should be prepared to do a lot of experimentation.

At TelCentris, for example, Bratt said he plans to explore ways to identify the top influencers in the unified communications and VoIP markets in order to engage them in beta-testing programs and other outreach efforts.

Kumar, too, plans to someday monitor social networks to better understand customer sentiment patterns and how one user influences another.

Said Kumar: “I wouldn’t say we have a full social media [analytics] program yet. [But] we’re evolving. We’ll get there.”

Dig Deeper on Customer analytics

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.