Text analytics tools can quickly mine seemingly limitless piles of text-based content to help users reveal important patterns and, sometimes, the very hutman emotions and motivations behind those patterns.
That’s why many experts think that text analytics tools, which have grown in both functionality and popularity in recent years, hold a great deal of promise for business users.
Experts say text analytics tools offer CIOs the prospect of finally getting a better handle on unstructured content. It offers product and marketing managers the opportunity to listen more carefully to the “voice of the customer.” And text analytics tools give business intelligence (BI) software users the chance to add a great deal more insight, context and, as a result, meaning to the reports they run.
But text analytics projects don’t always go smoothly. In-house text analytics software projects require expensive architecture changes and expertise in areas that probably aren’t familiar to typical IT personnel. Out-of-box text analytics software packages, meanwhile, come with their own set of challenges, which range from the technical to the cultural.
Businesses willing to work through the text analytics challenges are likely to be rewarded with a strong ROI in the end -- the result of savings on marketing costs and new revenue generated as companies do a better job of responding to customer sentiment.
SearchBusinessAnalytics.com recently spoke with a handful of text analytics experts at the Text Analytics Summit 2010 in Boston. Here, in no particular order, are the top five text analytics tips we heard that day.
Top five text analytics software tips
1. Use text analytics software to prevent problems
Text analytics tools give companies the ability to find information hidden in emails, on social media and other websites and in any number of electronic documents. Those capabilities often uncover issues that customers have with products or services, and those problems then require damage control. But text analytics can also be used to help solve problems before they arise or, at the very least, before they get out of hand.
Bahman Dehkordi, a research statistician with State Farm Insurance in Bloomington, Ill., said he wants to learn more about text analytics to see whether the act of incorporating textual data into his calculations could help prevent problems before they happen. Suppose State Farm were thinking about changing a popular insurance policy, Dehkordi explained. The statistician could potentially use text analytics, along with other calculations, to learn how people reacted to similar changes in the past. Dehkordi could then use that information to make an inference or a prediction about whether the proposed change would end up being good or bad for business.
“The message is to be preventative rather than trying to be reactive,” Dehkordi said. “The whole idea with [text analytics] and customer relationship management is: Can I think ahead? Can I use the information I have to predict what is going to happen to me?”
2. Use text analytics to examine many information sources
Chris Jones spends a lot of time working with -- and teaching people about -- text analytics tools. As manager of analytics of Intuit, the Mountain View, Calif.-based maker of well-known financial applications like TurboTax and Quicken, Jones said one problem he often sees is the tendency of people to use text analytics to study just one or very few data sources. To Jones, this is a waste of a valuable and highly versatile resource.
“I see a lot of people that will look at one source of data like Twitter or social media, and that is all they’ll look at,” he said.
Intuit’s analytics team uses text analytics tools to study myriad information sources. They include product reviews, product feedback, surveys, call center data, online communities, post contact surveys, information gleaned from focus groups, results from usability labs -- and the list goes on.
“People ask, ‘What is the voice of our customer?’ It really means all of these things,” Jones said. “It’s not just what people are tweeting about.”
3. Understand that with text analytics, there’s no such thing as 100% accuracy
IT folks who run text analytics software deployments are often tasked with bringing specific findings to business users who then use that information to do things like create new products or craft some sort of response to overall customer opinion.
One problem that IT folks repeatedly mentioned at the conference was that the business users they support often complain when text analytics-based searches for information provide non-definitive results, leaving everyone with an uncertain course of action.
Analysts and text analytics users say it’s important to make those business types understand that these situations come up more often than not, and they need to do the best they can with the information available. IT pros can help in this respect by using their knowledge of text analytics to “look at the big picture” and provide the business user with key areas of focus.
“They’re going to focus on the inaccuracy of some of the analysis … and [IT professionals need to] get them to understand what the value is in what they’re giving them,” said Sue Feldman, an analyst with Framingham, Mass.-based technology research firm IDC. “There is not a single product that is perfectly accurate, neither is there a person who is perfectly accurate.”
4. Create customer experience teams
When approaching a huge and seemingly insoluble problem, it’s best to break that problem down into more manageable components. Then attack it one component at a time. Companies that plan to use text analytics as part of an effort to gauge customer sentiment should take a similar tack. That means forming customer experience teams that are segmented by each product, class of product, or service that the company offers. It’s a methodology that appears to be working well for Jones and Intuit.
“What this does is it basically gives clarity around who owns what,” Jones said.
The teams at Intuit do more than just pore over the results of text analytics searches. They are responsible for understanding everything that goes into building positive customer sentiment around their specific products. Companies should hold regular meetings between individual teams and executives. And be sure to pick highly responsible team members, because when customer sentiment problems arise, company executives will know exactly who to hold accountable.
5. Prepare for eDiscovery
Jason R. Baron, Esq., the director of litigation with the U.S. National Archives and Records Administration, had a message for the Text Analytics Summit attendees: Sooner or later, corporations get sued, and IT professionals who can help in the eDiscovery process will be extremely valuable to the organization.
It may sound like more of a career tip than a text analytics tip, but by helping companies prepare for eDiscovery – tagging necessary documents, meeting with general counsel prior to major purchases to make sure documents are searchable and to head off any possible trouble – IT professionals are actually helping themselves.
“There is a world of lawyers waiting for each of you [IT professionals] to inform them how to do search and retrieval in the legal domain,” Baron said with a smile. “If you can just [use] that key I've given you, it's worth about a billion dollars for your organization.”