Insurance companies collect huge amounts of text-based information daily, in many languages and dialects: customer feedback, emails, Web documents, blogs, Twitter feeds, adjuster notes, medical records, police statements, surveys, research studies, underwriter notes, competitive intelligence and more. No one has time to read it all, much less classify the content into common themes or make sense of the essential information.
To observe trends, spot new topics, issue alerts about potential problems and flag new business indicators, insurers must be able to analyze all your data before acting on it. But conversational language is ambiguous, and key messages buried in text data are not easy to discern or process. Most organizations are unable to combine text content with structured data in decision-making contexts.
By most estimates, up to 90% of the information available in an organization is actually unstructured data, and that percentage is likely to increase with the growth of social media. But for insurance companies, it might just as well be called invisible data. For an industry that is driven by data, text analytics is still new to most insurance companies.
So what is text analytics? The simple definition is the use of computer software to annotate and extract information from electronic text sources and analyze that information for business purposes. But pattern discovery is where we get the true value of text analytics. Within this category the main technologies are:
• Sentiment analysis, which automatically locates and extracts sentiment from online materials.
• Text mining, which provides powerful ways to explore unstructured data to discover previously unknown concepts and patterns.
[Previously from Rose: Finding your most valuable policyholders]
Imagine the power if insurers could harness the insights hidden within that vast sea of words? Here are just three areas where text analytics can positively impact insurer profitability:
It is estimated that, on average, 5% of claims that should go to subrogation don't. By using data and text mining techniques, insurers have minimized the number of missed recovery cases by recognizing known and unknown subrogation indicators in the claims information. In fact, one leading European insurer was able to improve its recovery rate by more than 4%, representing millions of dollars per year to its bottom line.
Insurance companies are using text analytics to dig into the details contained in applications, adjuster notes, and other unstructured text sources, helping prioritize cases for SIU examiners. For example, carriers may find a common phrase or description of an accident among multiple claimants, usually an indicator of organized fraud.
Our world has changed drastically in the past 10 years. Before the rise of the Internet, blogs and social media sites, an insurer could more easily maintain and control its brand and reputation. Now these websites are creating a virtual focus group online. Customers are venting their frustration or sharing stories of positive experience, and all of it is recorded and shared with millions of other consumers. Sentiment analysis can analyze this online content and categorize the text as "positive," "negative" or "neutral." Insurers can then review the various categories and comments and look at trends over a period of time to see if their brand reputation is improving or declining.
The volume of unstructured data coming from social media is growing exponentially. Add to this the thousands of call center records, emails and other internal data sources. Today successful and innovative insurance companies can no longer ignore text analytics and the vast amount of information contained within unstructured data.
About the Author: Stuart Rose is global insurance marketing manager at Cary, N.C.-based SAS. Rose, a 20-year veteran of the insurance industry, began his career as an actuary. He has worked for a global insurance carrier in both its life and property divisions and has worked for several software vendors, where he was responsible for marketing, product management and application development. He has driven successful development and implementation of enterprise systems with insurance companies in the U.S., the U.K., South Africa and Continental Europe.