Mining Free-Form Data Enables Better Customer Service
Recognizing that virtually any part of the information a company takes in may be critical in delivering better value and customer service, vendors are developing new technologies that enable companies to mine common text data that might otherwise be ignored.
According to San Mateo, Calif.-based Enkata Technologies, organizations today can evaluate and implement technologies that can mine and analyze “free-form text documents,” including customer service representative (CSR) notes, customer support e-mails and chat logs.
Free form text data is any information that is captured in a written formincluding notes in a call center, e-mail between an agency and its customers, and claims adjuster notes, says Josh Klaher, director of product management for Enkata. Such text may also include system-to-system generated text.
“When a call center agent picks up a phone and checks on a customers claim status, that transaction may include lots of information,” he explains.
Uncovering such data “provides a lot more context around the nature of the interaction with the customer,” says Klaher. The information enables a company to focus on customer interactions, which include any activities related to customer issues. Companies can then examine customer interaction for time-wasting patterns or other factors that may cause problems, he explains.
“For example, if you get repeated calls on the same claim, text mining builds a picture of these defective transactions,” says Klaher. “You may have CSR notes or claims adjuster notes that describe information about the claim. The CSR may record a note that says, customer disagrees with deductible, and this information may not be anywhere else.”
According to Klaher, mining such data “definitely helps you know your customer better, and it reduces costs by helping you figure out where to improve processes to reduce the number of contacts with customers, or how to encourage customers to perform self service.”
Companies may discover that only a certain type of claim is getting hung up, or that duplicate claims are being filed, he notes. Overall, the technology can help a company focus on root cause drivers of these problems.
“When you include information that is stored in text documents, it helps paint a much better picture of interactions and may reveal the smoking gun,” says Klaher. “If the dispute deductible problem shows up a lot, maybe youre not doing a good job of explaining how that works.”
Fixing such problems, he adds, improves customer satisfaction.
From an analytic perspective, however, it can be “pretty challenging to get at this data,” he notes. “A disputed deductible concept may be represented many different ways, so a simple keyword search wont work, especially if deductible is spelled four different ways.
“Text data is often ugly,” he continues. “It takes a long time to churn through and put that data into an analytical schema.”
The ability to analyze such data, however, has been enabled by increases in commonly available computing power.
“Its not sufficient to analyze data in a vacuum, however,” Klaher emphasizes. Companies need to be able to merge their free-form data with already structured data on customers in order to provide valuable insights.
Technology solutions to mine and integrate such information can be built internally or purchased from software vendors, he explains. The key is to give companies a view of whats happening across their businesses, rather than simply recording a claim or a phone call.
“Were linking a huge number of dimensions to customer interactions, so one real challenge is how to visualize all the information once its merged,” says Klaher, whose company provides such solutions.
Enkata offers two main views in its application: a detailed view of all elements, or a summary view which uses a statistical “relevance score” that indicates how relevant data is to the problem youre examining.
For many companies, Klaher explains, a challenging question is: “How do I encourage my customer to use self service on my Web portal?” People use such portals, but the portals are often not satisfying all of a customers needs for interaction, he notes. “They still have to make a phone call to get the rest of the information.”
Using data mining technology to review those customer calls, companies can find out what is happening and what additional information to provide to avoid receiving calls, which can drive up costs.
According to Klaher, a companys investment to make such technology part of its systems will typically be in the $1 million range. That would include software and implementation, along with doing all the necessary preparatory work and integrating all of the tools used.
“The thing that impacts cost is volume of data and number of data sources,” he states.
Many customers will start with a smaller projecta pilot implementation with a limited number of data sources and focusing on a single problem. “No one wants to drop a million bucks on something that hasnt been proven to them,” he notes.
Enkata says it is providing a free guide to mining text data via its Web site. The guide is written for managers and executives of “large, customer service-intensive organizations,” such as insurance companies. Titled “An Introductory Guide to Mining Text Data,” the paper explains how companies can evaluate and implement technology that can mine and analyze free-form information.
The guide is available at www.enkata.com/contact/downloads.html.
Reproduced from National Underwriter Life & Health/Financial Services Edition, October 24, 2003. Copyright 2003 by The National Underwriter Company in the serial publication. All rights reserved.Copyright in this article as an independent work may be held by the author.