+91 96558 14047 (India)
+65 8237 9397 (Singapore)
+27 11 886 1707 (South Africa)
+61 8 4634 1736 (Australia)
+44 (0) 208 123 3459 (UK)
+1 315 532 7622 (USA)
Insurance claim fraud is one of the major avoidable losses that hurt insurers all over the world. Most of the fraudulent insurance claims are from the P&C segment with auto insurance and workers compensation accounting for the largest portion of fraudulent claims to impact the insurance industry every year. Traditionally, insurers use manual methods of fraud detection. These methods have their inherent disadvantages because of which undetected instances of fraud continue to bear heavily on insurers. When only sampling methods are used to analyse claims for frauds, some fraudulent claims slip through the system and as this method relies on historical fraud data, new frauds could easily go undetected as well. Added to this, as the traditional method works only in silos, it is not built to handle the influx of data and information from many new and different sources of information in an integrated way.
The 10 step approach for implementation of analytics in fraud detection
As many insurers realise the critical role played by analytics in fraud detection, many of them rush to implement expensive fraud solutions that may not be in line with the company’s strengths or weaknesses. It is essential therefore for insurers to do a complete SWOT analysis, identify critical touch-points and then purchase relevant fraud detection solutions in order to make the fullest use of analytical solutions.
Most insurance companies do not have specific teams or individuals whose task it is to detect fraudulent claims. Whenever a fraud is suspected or detected, people come together to combat it. For viable fraud detection, a specialised team needs to be put together, trained, equipped with the right tools and given a line of reporting to senior management to focus on efficiency and effectiveness.
On completion of the SWOT and once the dedicated team for fraud detection is in place, insurers need to think through their requirement for analytics in fraud detection and examine if they need to buy a solution or if they have the skill set available in-house to build it. If the decision is to buy, an evaluation of different vendors should be performed considering costs, user interface, scalability, ease of integration and ability to add new data sources.
Insurers will need to integrate databases that have hitherto been siloed and put in place a robust process to remove inefficiencies from processes and redundancies from all the sources of data. Adequate testing of the cleansed data needs to be done in order to ensure accuracy.
As the insurance industry and certain companies within the industry are prone to certain specific types of fraud, insurers must utilise in-house experience, domain expertise and experienced resources to establish a framework of rules to build an efficient fraud detection system.
Irrespective of whether the analytics framework is built externally or internally, insurers must provide for threshold values for different types of anomalies. Setting precise thresholds for the different anomalies is the key to an effective fraud detection system. If the threshold is set too high, many fraudulent claims could slip through the system. And if the threshold is too low, it could result in too many claims held up, time wasted, members and service providers losing patience and even late-payment penalties.
Social Media Analysis (SNA) is an effective method of fraud detection. It helps to identify organised fraudulent activities by closely examining the relationships between the different entities involved in a claim. Entities can be as diverse as addresses to telephone numbers. The number of linkages found to each entity when compared to the average number of connections that can be expected will throw up variances that can point to organised fraudulent activity.
Integrated case management calls for fraud investigators to capture all key findings and compile a thorough profile of the case in question. Key findings include claims data, network diagrams, adjuster notes and data from social media and can be in a structured or an unstructured form. These indicators, once tabulated are compared with standard metrics to determine if the case is one of fraud or abuse. Case management and workflows enable complete analysis of the investigative workload, efficiencies and the returns on investment.
The implementation of a fraud detection system does not stop once it is initially set up. Insurers will need to proactively look for additional sources of data input that will augment their existing fraud detection mechanisms in order to build a system that will continuously evolve and be able to detect new classes of fraud that may arise in the future.
With specialised solutions for the insurance industry, Neutrinos is helping leading insurers take impactful digital transformation decisions. Neutrinos is in a position to deliver value through a results-based approach, tailored to your specific goals.
Reach out to our experts for more details.