Text Analytics: Your Customers Are Talking About You
- 17 June, 2008 08:17
If you're in the hotel business, customer satisfaction isn't just a key metric, it's one that can make or break the company. But until recently, addressing sources of customer dissatisfaction was taking too long for Gaylord Hotels. US-based Gaylord, which operates 4 resort hotels in the Nashville; Dallas; Orlando, Florida; and Washington areas, needed a quick, clear view of how customers and meeting planners viewed its properties and services, as well as alerts to budding problems.
"Our survey vendor would do manual categorization, essentially reading [customer] comments and getting back to us," says Tony Bodoh, Gaylord's operations analysis manager. In search of faster results, Gaylord turned to text analytics technology from Clarabridge, beginning with a pilot test in 2007 followed by a phased deployment in 2008.
Text analytics, often referred to as "voice of the customer technology," is designed to squeeze sentiment out of customer communications rather than simply retrieve isolated nuggets of information, as traditional text mining does.
"One of the key benefits the Clarabridge tool provides is essentially overnight categorization and clustering of all the comments," Bodoh says, "which was taking us several weeks to a month with the previous vendor."
Bodoh says the technology is already beginning to help the company pinpoint specific sources of guest dissatisfaction. "One property may use a different vendor for purchasing a particular product," Bodoh says. Viewing guest comments on topics such as bathroom cleanliness or restaurant service helps Gaylord managers spot weak performers, he says. "We are also using the software to understand best practices across our hotels, and how to bring those best practices from one hotel to another hotel, or from one department to another department," he adds.
Clarabridge, along with Attensity, Business Objects, SAS and several other vendors, offers software designed to help enterprises understand and learn from what customers are saying about products and services. Along with surveys, e-mail and phone calls, the technology can monitor blogs, text messages, online chats, phone calls (through speech-to-text conversion) and social network profiles.
While text analytics today is far from an out-of-the-box solution, CIOs say, the technology may give you insight into customer thinking that's hard to put a price on.
In Search of Trouble
If your company allows customers to talk about products and services on the company website, for example, text analytics tools can help you analyze what those comments and chats say, to improve business decisions and strategy.
"Just about anything that's in text or can be converted into text," can be analyzed, says Matthew Brown, principal analyst for information and knowledge management at Forrester Research. Businesses in diverse fields including transportation, hospitality, business and consumer products, retail, entertainment and even law are beginning to embrace text analytics, Brown notes.
Text analytics tools also enable an enterprise to scour the digital grapevine to pinpoint budding problems that could tarnish a brand's luster, says Fern Halper, a partner at Hurwitz & Associates, a consulting and research firm. "The software helps companies understand what customers are saying about their brands, so they can actually get a head start in finding problems before they occur and make course corrections in midstream," Halper says.
For Internet travel giant Travelocity, that desire is keen, says Ginny Mahl, Travelocity's VP of customer care and sales. And every day, customers send the company plenty of content to examine, Mahl says. From North America alone, Travelocity each month receives some 25,000 to 30,000 customer satisfaction survey responses, 35,000 to 50,000 e-mails and 400,000 calls, she says.
To help sort through this digital haystack for insight needles, Travelocity turned to Attensity's text analytics tools. At press time, Travelocity was preparing to deploy a production version of the software. "We are using it primarily to read verbatim feedback from our customers to gain insight into likes and dislikes about Travelocity, and recommendations they have for improvements in our products and services," Mahl says.
"In particular, the application will let us do a much deeper dive into the root causes/drivers of the satisfaction scores we receive."
Mahl expects that the software will enable Travelocity to detect consumer sentiment trends that may impact customer satisfaction. Mahl offers an example: "Through a very simple query, we're able to 'read' our customer comments and find out if the amenity offerings at a particular hotel have changed, allowing us to update our website content more rapidly," she says. "This capability is one of those very simple things that can have a big impact on our customers' experience."
Text analytics also promises to help Travelocity bolster its partner relationships, Mahl says. "We [receive] supplier-specific feedback, which we can feed back to our partners so they also have a better understanding of our mutual customers' opinions," she says. "Sharing valuable customer information is in both of our best interests."
Travelocity, which first learned about text analytics from vendors who approached the company, evaluated several products, Mahl says. Two factors drove Travelocity toward Attensity, she says. "Attensity's solution lets us acquire an enterprise license for the software, which we can scale much more economically as we grow our application of text analytics to additional data or new data sets," Mahl says. "They were also able to provide a travel industry taxonomy to jump-start analysis."
Rolled-Up Sleeves Needed
At Gaylord Hotels, the Clarabridge tools are helping the company address a variety of emerging problems quickly, Bodoh says.
For example, the software recently confirmed for Gaylord managers a problem synchronizing its automatic door locks, preventing key holders from opening room doors.
"For a few days, we were seeing a spike in the number of complaints around these key problems," Bodoh says.
As it turned out, the annual changeover from Daylight Savings Time to Standard Time was the culprit.
"We identified that it was related to the changing of the time in the software that runs the locking system," he says. "That's an example of something that we normally wouldn't have seen, but all of a sudden it popped up to the top."
"Before text analytics arrived, enterprises were limited to manually analyzing a small percentage of the unstructured customer information they received," says Sid Banerjee, Calrabridge's CEO and chairman.
But while text analytics can rapidly generate vast amounts of deep customer insight, the technology is still far away from becoming an out-of-the-box solution, IT leaders say.
"I'd consider it complex," Travelocity's Mahl says. "For our purposes, in order to get full value from the application, we will have to train analysts to use the software, invest in tuning the taxonomy to produce more granular analyses and integrate the output...with our enterprise data warehouse so we can use the combined data for even greater customer insight."
Mahl would also like to see Attensity extend its software to include real-time access to additional data sources, such as websites and blogs, both of which would allow Travelocity to monitor even more customer views.
Gaylord's Bodoh notes that many of the text analytics applications that he's examined support only a limited range of analytic topics. "Some would only be able to categorize 40 or 50 problems," he says. "With the categorization model that we have in place, we have over 300 categories."
IT leaders examining these tools should also consider the number of "sentiment" words that an application can recognize, Bodoh warns.
"Clarabridge has the capability to really comprehend what the sentiment is for about 25,000 to 30,000 words, and you can tweak that for your industry," Bodoh notes.
As a growing number or enterprises adopt text analytics, the technology is likely to migrate into other business applications that brush up against users' thoughts and opinions.
For content management products, Halper notes, text analytics can be a complementary technology; for example, text analytics can help categorize or enrich content, analyze content in a data repository, or improve workflow.
Also, vertical industries, such as the legal industry, are becoming increasingly intrigued by text analytics' ability to add insight to an array of routine business tasks, she says.
Today, many text analytics users believe that the technology provides a useful bridge to help nontechnical staff members get a handle on complex problems without running high-level searches.
"You can have business users who are not analysts really understand 'What are my top 10 problems?' or 'How is this issue trending over time?'" Bodoh says.
The software makes it easier for just about any business to maintain customer service standards while operating at maximum efficiency, Bodoh believes. "Managers don't have to worry about having to do a ton of analytics," he says. "It's very easy for them to understand."
Mahl, meanwhile, says she's satisfied with the results she's seen so far. "Although we are just getting started with the production implementation, we have seen nothing that changes our opinion," she says.
One warning: Do understand that these are often young and complex software packages. If your enterprise plans to evaluate the technology, test it with real consumer data, Mahl advises.
"Go through a thorough analysis of the software and, if possible, a live pilot focused on real business use cases in your company to demonstrate that the software is suitable for the intended use in your environment," Mahl says.