Analytics isn't a new concept, but considering the way insurers now use it, it may as well be. Where once fraud was the issue to enter insurers' minds when they heard the term analytics, they have now found more ways to use this technology. One approach is to employ text analytics, which often uses natural language processing (NLP), a subfield of artificial intelligence and computational linguistics, to convert unstructured data into valuable, structured data.
Frank Brooks, senior manager of data resource management at BlueCross BlueShield of Tennessee Inc. (BCBST), Chattanooga, defines text analytics as a technology that can transform unstructured text embedded in documents and databases into meaningful information in the form of new structured database fields.
Text analytics is vital to BCBST's enterprise content management strategy, which includes combing structured and unstructured data for analysis in order to provide valuable new insight into the business.
"Our initial challenge was our desire to know everything about a provider, which entails several things," Brooks says. "We started to see applications that need access not only to internal structured data, but access to external data and unstructured data, and we need to derive meaning or insight from the unstructured data."
BCBST evaluated two different text-analyzing tools during a proof-of-concept product evaluation earlier this year. Text Miner, a tool from Cary, N.C.-based SAS Institute Inc., is a complex solution that provides deep analysis of data and enables newly derived, structured data to be analyzed by existing BI tools. A second offering-IBM's OmniFind Analytics Edition-analyzes business language and is better suited for business-oriented people across the enterprise, Brooks says. "We've found if the user is in a specialized industry, two tools may be required-one for general business things such as customer service or contracts, and the other for specific knowledge based on the language used in that industry," he says.
The Hartford recognizes the technology's possibilities. "We use analytics for both structured data and recently, unstructured data," says Kaleb Adams, assistant VP, claims research, The Hartford Financial Services Group Inc., Hartford, Conn. "I've really tried to push text analytics out to the broader aspects of the organization."
Adams points out the fact that insurers are collecting more data every day and much of that data is not usable. "I read that 85% of data in claims was in unstructured text. We have claims that are still on the books from the 1950s. Insurers capture tons of data and text analytics mines that text and translates it into structured data."
Mary Crissey, analytics marketing manager at SAS agrees that insurers are collecting data they aren't using. "Insurers have been getting their data warehouses set up to keep and hold all those data sets and are collecting so much data they can't manually read it, so it's just sitting there." But this information can be helpful if it's turned into insight, she says. Text analytics will pull, clean, and analyze data. "It can conduct pattern detecting and cluster algorithms," Crissey says.
USE WITH BI
This new data can be used in many ways - including in conjunction with business intelligence. Text analytics tools can generate a structured data set that will interface with BI tools. "The BI field has been growing and continues to emerge and improve, and insurers have always been crunching numbers," Crissey says. "They're great with numbers, metrics, dollars and rates; that's their world, but they are still not taking full advantage of text."
Text analytics can turn the data and text that a computer can't structure into data with meaning, so it can be used in BI applications. "BI is based on having structured information," says Rita Knox analyst and research vice president at Stamford, Conn.-based Gartner Inc. "You have it in some sort of database where you know exactly what that piece of content is. It's in rows and columns, which you can view by drilling down to many levels. Text analytics identifies unstructured data in a way that enables BI applications to process the information."
Using text analytics can provide information and data that a user isn't even expecting, which is what The Hartford found when searching for a text analytics solution. "We brought in a series of vendors and crafted business cases for them. We gave them a series of note documents, sent them away and told them to come back with what they could find," Adams says. "We were able to give them a real business problem and evaluate each of the text mining solutions on the merits of their performance versus the merits of their Power Point presentation."
The Hartford settled on a solution provided by a Palo Alto, Calif.-based text analytics software provider. The solution, which uses NLP, was able to "mine the data not only for the obvious things but insights that we really weren't expecting," Adams says. "The text analytics space is fairly limitless with the opportunities because it can provide so much data. It goes well beyond what you capture in a structured data field."
OTHER USES
Customer service benefits derived from text analytics include the opportunity for upselling, says SAS's Crissey. "Somebody could be calling in to ask about their insurance policy and you could type in or record what the person is saying, and it could prompt a call center person to take an action, such as offering the person a certain price special at that time."
Knox suggests using text analytics to gain insight into customer opinion. "Clients call into insurers' call centers, or perhaps connect with the carrier via e-mails-these multiple sources contain information that is relevant to the company. It'll come down, in many cases, to text analytics being the core application. But on top of that, it enables the carrier to look for terms, for example, that indicate a customer's liking something or disliking something," she says.
On the other end of customer service, insurers can use text analytics to monitor the quality and effectiveness of the reps taking the phone calls. "You can analyze the results of their phone calls," he says. "There is a lot of data we collect in the form of hand-written or typed notes, voice files, etc. that need to be re-engineered into an electronic text format. Previously the data wasn't of much value, so we didn't want to take the time to convert it to usable data. Now, many of our business processes that capture unstructured data may have to be reconsidered."
One such business process is claims. Conversion and analysis of unstructured claims data can be used to improve and streamline the claims process. Yet, there is more insurers can do with the data, according to Crissey. "Processed claims have words and descriptive text in them," she says. "You can take care of that particular claim, but you can also use that data to do strategic thinking and planning."
Indeed, Adams says The Hartford uses analyzed text when developing new products.
BCBST is looking ahead to other areas to apply text analytics. "We're working with users right now, during our budget season, to see what projects we'll be doing next year and identifying the projects that could benefit from text analytics," Brooks says. "IBM announced on October 15 that the latest version of DB2-Viper 2 version 9.5-will have a text analytics capability. We plan to evaluate this over the coming months."
Brooks and Adams agree that text analytics technology will continue to improve. "We're going to see a maturing of the technology, but we're also going to see a maturing of the unstructured data," Brooks says. "We're going to have better standards for words, abbreviations and syntax."
And, the result of that improvement has-and will-benefit the business. "Before text analytics' evolvement, part of the problem was that you could do very only basic things, such as a keyword search," Adams says. "But, we're coming to a state where we're able to do more sophisticated research with more resources and better technology."
