Text analytics: versatile and growing

Using Verint, a payment processing provider built a predictive category for customers at risk, and a small team of agents reviewed the list each day. From that group, the agents identified the high-value customers and contacted them, saving 86 percent of the customers they contacted. In contrast, 98 percent of customers who were not contacted terminated their service. "Given the high cost of obtaining new customers, this action saved the company an estimated $12 million in the first year," he says.

Verint's solutions mine audio, video and text, and distribute the information across different departments such as customer relationship management (CRM) and marketing. "It's important to provide the feedback to the appropriate department, so they can make adjustments," Ziv says. Taking action on that feedback is the key ingredient for achieving ROI from voice of the customer initiatives.

Finding a direction

An important determinant of success in using text analytics is having a solid team in place. "The team should have a clear vision not only in how the application is going to be implemented," says Melissa Pippine, VP of marketing at Clarabridge, "but also about why the company is doing it and how the plan will be communicated internally."

The increasing recognition that new customers are hard to get is a strong motivator to use the feedback from text analytics. "With the decline in print and direct mailing, it is harder to get customers' eyes on your product, and finding new ways of reaching them is difficult and costly," Pippine explains. "Employees throughout the company need to fully understand that if they don't listen to their customers, their competitors will."

Some companies, especially those in retail and telecommunications, are analyzing Facebook postings and Tweets (twitter.com) that address problems their competitors are having, according to Pippine. "The companies then build marketing campaigns around those issues," she adds.

Clarabridge's products focus on customer experience management. The product suite includes an analytics solution as well as a collaboration capability that allows input from multiple employees for complex problem solving. "Clarabridge 6.0 includes new reporting capabilities for stakeholders throughout the enterprise and also support for Hadoop to help address and productively use the endless amount of data now being generated," Pippine says.

From the mundane to the fraudulent

Not all applications of text analytics are highly sophisticated, but they can still add significant value. "One company had developed a large taxonomy but it had not updated in 10 years," says Tom Reamy, chief knowledge architect at KAPS Group, which specializes in knowledge architecture. "We used text analytics to discover the most common new terms emerging, so new categories could be developed and metadata assigned."

Another straightforward application is document de-duplication. "Most companies have 30 different versions of documents floating around in their content management systems. Trying to find the right one is time-consuming and expensive," Reamy adds. "Text analytics can be used to identify the right version, and huge cost savings accrue from removing the unnecessary ones." Search accounts for about 30 percent of the market for text analytics, according to Reamy.

Business intelligence tools have long been used to detect patterns in structured data that indicate fraud; text analytics is increasingly being used for the same purpose. E-mails from the Enron case provide a set of materials for the study of how language can indicate fraud. "It turns out that people use words differently when they are lying versus when they are telling the truth," Reamy explains. "The differences in word patterns are fairly striking, and analysts are achieving an accuracy of about 75 percent in detecting deception."

"Text analytics went through one standard hype cycle with sentiment analysis and voice of the customer applications," Reamy says, "but then companies started looking more carefully at the business value and about how it should be used."

Barriers still exist because it is a relatively new technology that can be somewhat difficult to develop and use. In addition, advocates need to come up with a quick win, or companies do not want to invest additional resources. Still, Reamy estimates that the market is growing about 20 to 25 percent per year, indicating that the need for processing unstructured content is not going to abate any time soon.