If you’re just catching on to the fact that big data is shaking up the insurance industry in a big way, you’re a bit behind the curve.
Across all industries, almost 90 percent of large companies say big data is going to revolutionize business operations, according to a report from Accenture. It’s predicted that this revolution will be on a scale comparable with how the internet changed the way we work starting in the 1990s. Remember the way it was before the days of email, Google and filing claims online?
Understandably, companies are scrambling to identify the new realities big data will bring, and seize the opportunity it presents. According to that same Accenture survey, 83 percent of companies plan to pursue big data projects in order to try to gain a competitive edge.
We have already seen an inkling of big data’s impact, especially in claims departments that are increasingly finding themselves on the front lines of these developments. From telematics to text mining, experts see data analytics significantly reshaping the claims process in areas like fraud detection, claims triage and process efficiency.
But organizations looking to successfully develop and implement new data-driven processes still face considerable hurdles, especially when it comes to cost. Collecting, analyzing and even storing all the data insurers collect can be prohibitively expensive.
Currently, insurers only process about 10 to 15 percent of the structured data they have at their disposal. That structured data, which is already sorted and organized in databases, is easier and cheaper to manipulate than unstructured data. The trouble is that an estimated 80 percent of all data businesses use is unstructured.
In the claims world that figure may be even higher. Police reports, witness accounts, claimant statements and other adjuster notes all need to be processed and cleaned up (data pros call it scrubbing) before it can be put to actionable use.
The Power of Text Mining
Data scientists and insurance professionals are turning to text mining as a relatively cost-effective way of uncovering the value hidden in these mountains of unstructured data.
Text mining encompasses extracting useful or interesting information from unstructured text. In its most basic form, text mining involves scanning large amounts of data for keywords or phrases, similar to a Google search.
But thanks to artificial intelligence advancements such as natural language processing and decision logic, modern text mining goes beyond simply finding information to analyzing the documents for significant facts and relationships within the data.
Today, text mining is capable of scanning reports and interpreting adjusters’ handwriting. It can discover customers’ sentiments, from their opinion of a product to how they’re feeling immediately after filing a claim. The sources of this data are increasingly expanding beyond information directly collected during the claims process. Rather than having an adjuster sift through a claimant’s Facebook or Twitter profile, a text mining algorithm can search all of the person’s social media content in real time for information that contradicts or confirms specific details of a claim.
Streamlining the Claims Process
These text mining tools have the potential to significantly increase efficiency within the claims department. Analyzing claim filing call transcripts can reveal better ways of structuring and scheduling call center operations. Identifying keywords used in reports that indicate a high potential for a complicated claim, allegations of bad faith or litigation could get special attention or be assigned to a more experienced adjuster earlier in the process.
Text mining of claims data can also be an extremely useful tool in developing new products and plugging gaps in existing coverage that may be negatively affecting customer satisfaction and renewal rates. It can also create more effective competitive analyses and drill down into demographics to identify where to focus marketing efforts to attract more customers.
Rooting Out Fraud
Currently, many insurance organizations are using text mining most effectively to identify fraud potential.
As of 2014, 43 percent of insurers were using text mining in fraud-fighting efforts, according to SAS. In some cases, text analysis and other big data techniques allow insurers to identify applicants who are more likely to commit fraud at the underwriting stage. But a majority of fraud detection occurs in the claims process.
Text mining can spot red flags that multiple claims adjusters may miss or never even be exposed to. If, for example, several claimants are using the exact same language in filing a claim, a text analysis will mark that suspicious activity. An adjuster may not notice the phrases repeated verbatim, or multiple adjusters may be assigned to those cases and have no chance to notice the scripted phrases. Text mining can catch this. It can also quickly identify physicians, body shops and even witnesses whose involvement in seemingly unrelated claims raises concerns about fraud.
Where Claim and Data Meet
Data scientists excel at creating algorithms that extract text data efficiently and present the results in a revelatory and meaningful way.
But data scientists are not claims experts. Without guidance, there is no way for them to know the keywords to build into their algorithms or what revelations will be the most valuable to the claims department.
As insurance organizations develop the data analytics tools that help streamline processes, identify fraud and reveal opportunities for new products, claims professionals and managers must have a seat at the big data table.
That means giving claims a prominent place on the teams responsible for developing and implementing data analytics and modeling processes. Adjusters and claims managers should help data scientists understand the specifics of how the claims process works at their organization and how analytics can help.
But the intersection of claims and data science is a two-way street. As data scientists work to develop a better understanding of the claims department, claims professionals should develop a basic understanding of data analytics and how it applies to the insurance industry. Understanding how to think in terms of data and how access to information can create opportunities and help solve problems.
Claims Expertise Isn’t Going Away
While text mining and other analytics and predictive modeling tools are poised to significantly disrupt fundamental insurance industry processes, do not expect the need for experienced claims professionals to disappear anytime soon. Just as data scientists and claims experts need to collaborate to create the algorithms that identify meaningful trends in text mining, insurers and customers will still rely on humans within the claims department to triage and tackle complicated claims.
Organizations that are able to incorporate meaningful analyses of unstructured data through techniques like text mining into their claims department will find themselves one step ahead of competitors with less of a grasp on big data’s potential.
But to stay competitive, claims departments will have to coordinate closely with data scientists to continue to refine processes and pull out even more value from the massive amounts of unstructured data just waiting to be analyzed.
Read another article by the authors: The Future of Claims Careers: Cultivating a Data Mindset
Judith Vaughan, CPCU, AIC, AIM, has more than twenty years’ experience in claims and risk management for international organizations, and is currently director of Content Development at The Institutes.