The Wall Street Journal-20080128-Business Solutions- So Many- Many Words- Companies have a treasure trove of customer intelligence- Now they also have the tools to make sense of it
Return to: The_Wall_Street_Journal-20080128
Business Solutions; So Many, Many Words: Companies have a treasure trove of customer intelligence; Now they also have the tools to make sense of it
Full Text (1314 words)A couple of years ago, Travelocity decided it had to get to know its customers better.
A company team spent several months poring over some 10,000 customer surveys, trying to figure out what people liked and didn't like about Travelocity. "As you can imagine, it was a very challenging process," says Don Hill, director of customer advocacy.
That was only the tip of the iceberg. To keep up with customers, Mr. Hill figured, his team would need to track and analyze 30,000 survey responses, 50,000 emails, and notes from half a million calls to the company's service centers -- every month.
The solution? Travelocity turned to software maker Attensity Corp., whose products quickly analyze documents and pull out vital information. Using the software, Travelocity pinpointed critical customer concerns and came up with fixes. For instance, Travelocity found that some clients held the company accountable when airlines cancelled flights. Now Travelocity is trying to help customers plan itineraries that are less likely to get hit with cancellations. It's also developing better methods to help customers whose flights are cancelled.
Over the past few years, many big organizations have found themselves facing the same problem as Travelocity. They have access to treasure troves of customer intelligence -- everything from surveys to emails to online reviews -- but can't sift through them effectively. Employees don't have the time to read each document individually, and traditional database programs are designed to handle numerical data, not words.
That's where software like Attensity's comes in. These programs -- known as text analytics -- can scrutinize text documents, quickly identify crucial terms and concepts and put all the information in an easily searchable form.
U.S. security agencies have been the biggest users of these technologies in the wake of 9/11. But a growing number of corporations are now discovering that text analytics can help them identify market trends and customer patterns, spot fraud and security threats, and highlight problems with products.
For instance, auto makers search through customer complaints, accident reports and insurance claims to quickly identify defects in their cars. One car company was getting positive feedback about one of its models in surveys and customer emails. But software analysis showed that customers were burying a complaint in their notes: There was an annoying squeaking sound in the back passenger area. The auto maker fixed the problem before it hurt the company's reputation.
How does text analytics work? The software -- programmed with dictionaries and a knowledge of grammar -- starts by identifying the different components in sentences, such as nouns and verbs, and subjects and objects. Using that information, it determines the major themes in each sentence and paragraph, as well as the overall document, and then figures out the relationships between them. Finally, the software stores all that information in a database, where users can search by key terms. A company might check to see how customers regarded its Web site's "ease of use," for instance.
It's unclear just how big -- or small -- the text-analytics market is at this point. The technology has been around for about a decade, and several industry insiders suggest that world-wide sales were a mere $250 million in 2007. (Final numbers weren't available.) But they also say the practice of analyzing documents with text-analytics software -- a process known as text mining -- has only really started to catch on in the past year or two. And they believe the sector is poised to grow 25% annually over the next several years.
A couple of factors are driving growth. First, the technology finally has reached the point where it can make sense of huge amounts of data. Computers have gotten more powerful in recent years, and vendors are harnessing that power more effectively, by developing specialized software for different industries and tasks.
That's part of the reason Travelocity, a unit of Sabre Holdings Corp. of Southlake, Texas, turned to Attensity: The Palo Alto, Calif., start-up had created a product geared toward the travel sector. Similarly, ClearForest, a Waltham, Mass., text-analytics start-up recently acquired by Reuters Group PLC of London, has developed products for the pharmaceutical and financial-services sectors. Other vendors focus on specific tasks. For example, Lexalytics Inc., of Amherst, Mass., offers software that helps clients track and evaluate comments on blogs.
Another big factor driving text analytics is the explosion in online opinion writing, such as blogs and review sites. Many companies now find themselves facing virtual gold mines of information that are tough to analyze manually. At the same time, the boom in these online outlets has made customers more demanding. Since people can easily share their opinions, they expect their voices will be heard.
"The market has changed. Customers can get what they want, and they are more vocal about what they like," says John Griggs, director of customer experience at H&R Block Inc., of Kansas City, Mo.
H&R Block plunged into text analytics two years ago, when it faced stiff competition for its do-it-yourself online tax-preparation service.
As a first step, H&R Block wanted to understand why some customers were willing to recommend the service to friends and others weren't. With much of its business compressed into the weeks around tax time, the company needed to respond quickly to promote features or services that were popular -- or fix problems that were turning off clients.
H&R Block turned to Clarabridge Inc., of Reston, Va., for software that analyzes surveys, customer emails and notes from calls to H&R Block service centers, providing early-morning reports on some 10,000 customer contacts from the previous day. Mr. Griggs says it used to take 67 hours to review 1,500 customer surveys by hand. Clarabridge's software can do it in 30 minutes.
H&R Block zeroed in on customers' problems with the online filing process, which might make them hesitant to recommend the service. For example, customers were confused about the 2006 Telephone Excise Tax Refund. More than 160 million filers were eligible to request it, but H&R Block discovered that many clients didn't realize they qualified. The company quickly altered its online forms, clarifying who was eligible and helping customers maximize their refunds.
Still, text analytics carries some risks. Vendors and users caution that companies looking to adopt the technology should consider some key points.
First, prospective buyers must have a clear sense of their goals, because text-mining products are designed for specific industries or tasks. Companies that roll out the wrong software could waste valuable time and money. And, perhaps most important, they could end up with incomplete or inaccurate data. For example, an electronics maker that is losing customers might decide to analyze surveys to figure out why people are unhappy. But the software it chooses might not monitor blogs, where consumers increasingly post their complaints. So the manufacturer might miss crucial information about why customers are dissatisfied.
Moreover, experts say companies should pick a vendor that has a track record in their industry. Often, that doesn't take a lot of homework; with the field still in its infancy, vendors are trying to establish their reputations, so they often tout testimonials and media reports on their Web sites. Also, in such a new industry, prospective customers should make sure a vendor can deliver on its promises. Both Travelocity and H&R Block insisted that their vendors set up pilot projects that could crunch real data in real time.
Travelocity and H&R Block say their initiatives are too new for them to measure the effect on the bottom line. Still, they're confident that better customer intelligence will lead to improved service, happier customers and higher revenue.
"Millions of our customers want to talk to us about our products, and text mining will help us hear what they are saying," says H&R Block's Mr. Griggs.
---
Mr. Morrison is a reporter for Dow Jones Newswires in San Francisco. He can be reached at [email protected].
