North American insurance companies are increasingly investing in analytics and “big data” to seek competitive advantages, assess risk more sensitively and combat fraud. For claims professionals, this may mean learning new processes and acquiring new insights.
What’s the big deal?
Many billions of bytes of data are created around the world every day. Some of this data comes from comparatively traditional sources such as databases; and other data – increasing quantities of it – comes from non-traditional sources such as online search records, blogs, customer comments on websites, messaging via mobile devices, and other social media.
“Big data” is any type of data that flows in large volumes – structured or unstructured data; text, audio or video files; sensor data, click streams, log files and more. The volume and diversity of big data make it hard for businesses to navigate and assess. Yet new insights emerge when these disparate data types are analyzed together, potentially allowing a firm to improve its competitiveness, efficiency and profitability.
This potentially creates a huge market for big-data technology and services. Some insurers are already investing significant sums in analytics, and others have plans to do so. Advanced analytical techniques can identify meaningful patterns and highlight emerging themes by very quickly processing big data.
Mining and modelling
In data mining and predictive modelling, historical data is carefully examined to identify patterns. Predictive modelling solutions can process an ever-growing list of risk characteristics in order to more effectively segment and price business, especially in commercial p&c lines.
Rating plans using predictive analytics can assign a unique price to any risk, making pricing a more potent underwriting tool. This allows an insurer to better manage risks in lines of business like business owner’s policy, commercial package, commercial property, and directors’ and officers’ liability by identifying policies that are likely to incur losses.
On the claims side, many companies are looking at their costs and asking whether they’re spending more than they should to settle claims, or whether there are specific claims that could be handled more efficiently. Predictive modelling can improve a company’s workflow efficiencies by providing claims handlers with timelier loss insights, allowing handlers to better focus on the right claims at the right times.
Analysis techniques can identify patterns in the data that would be difficult for adjusters to spot because of the vast amount of information involved and the limited time available. Proper use of predictive modelling can facilitate the early identification of potentially costly claims and the recognition of claims practices that are unnecessarily increasing payments.
Another area of claims cost leakage is subrogation. Carriers often have only partial information concerning the circumstances of a loss, the degree of negligence across involved parties, and the extent to which recovery is available or is pursued against another carrier. Improved data can help.
The use of predictive analytics also has significant implications in the fight against fraud. Fraudsters are very adept at masking their identities, but technological tools can examine hundreds of elements across millions of records to provide real-time alerts.
Fraud prevention is especially effective at two key points: underwriting and claims intake. For example, if the named insured or vehicle has been involved in previous claims, technology tools can identify which policies are being purchased for the purpose of conducting a staged accident.
The same technology can also confirm a person’s identity, the relationships between claimants, and the activities and names of individuals and businesses involved in the claim. Armed with these insights about the individuals, adjusters can change the questions they ask of claimants. A legitimate claimant will consider probing questions part of the normal process, but a fraudulent claimant may be exposed under the pressure of questioning if they feel they’re suspected.
Detecting fraud is a time-sensitive process, so a company needs to be able to analyze data as it comes in so as to maximize its value. And where data is used to support critical decision-making, establishing the reliability of the information is critical.
For predictive models to function effectively, they must be developed using the company’s own historical claims data, since this allows the model to recognize the specific nature of a company’s exposure and its claims practices. The process of developing such a model reveals the strengths and weaknesses of an organization’s document and record management practices, which are critical for ensuring the reliability of the data.
Integrating social media
Despite the excitement about the potential value of social data, use of this information is still in its formative stages. Integration of social data into core underwriting and claims processes remains a significant hurdle.
Authentication methods, data extraction tools and advanced analysis tools are just some of the techniques that must be developed or improved before social data can become standard inputs into risk evaluation and claims settlement activities. Insurers will need to investigate the best ways to integrate this data into their existing process and automation environments.
A similar integration hurdle exists at the claims investigation stage. While there are stories of insurers uncovering claims fraud by reviewing claimants’ social media websites, such practices have not become standard procedure. Moreover, claims searches are still manual and are heavily reliant on the experience and creativity of the individual investigator.
As with any new process, claims staff needs to be prepared for the implementation of new technologies and supported through the introduction. Change management processes involving all relevant staff, including individual claims handlers, may need to be implemented to ensure that the technology is not just adopted but routinely embedded in each adjuster’s practice.
In addition to the operational and technical challenges, companies need to keep sight of the legal issues connected with use of social data, especially privacy and consent. Culling information from social networking sites such as Facebook, Twitter and blogs may potentially infringe federal or provincial privacy legislation. Under certain circumstances, though, companies can collect and use personal information that has appeared in a publication that is available to the public (including a publication in electronic form) where the individual has provided the information. This is an evolving area of the law.
Technology often promises to make business easier and more efficient. For these promises to be fulfilled, change must be well managed so that the technology is adopted and embedded in routine practice. In the case of analytics, an organization’s records management system must also be strong, so that information can be catalogued and retrieved at the appropriate time; and integration processes will need to be established. Finally, privacy and consent issues will remain front-and-centre for some time as governments, companies and individuals adjust to the new possibilities.
This article is based on excerpts from ADVANTAGE Monthly, a series of topical papers on emerging trends and issues provided to members of the CIP Society. The Chartered Insurance Professionals’ (CIP) Society is the professional organization representing more than 16,000 graduates of the Insurance Institute’s Fellow Chartered Insurance Professional (FCIP) and Chartered Insurance Professional (CIP) programs.