1. Data aggregation and network analytics.
Traditional approaches use rules engines, matching techniques, or predictive analysis at the claim level only. Next-generation tools aggregate all historic policy and claims data into a data warehouse of networks, and apply all three of these techniques, in real-time or batch, to the claim, single historic customer view, and network.

Is a vehicle accident with three whiplashes fraudulent? Possibly. What if that policy is connected to others by key pieces of physical or behavioral data, and each of those other policies have had claims in the last few months involving multiple neck injuries? It is almost undoubtedly fraud. Conversely, following this exhaustive linking process, a claim that is on a network where a number of customers have had policies with the insurer for several years without claims has a higher probability of being a genuine accident. By networking all of your data, you have supreme intelligence to significantly improve the automated identification of fraudulent or genuine claims.

2. Multiple lines of business.
A number of insurers have demonstrated that building networks of customers and activities across multiple lines of business provides even further customer insight and intelligence. A customer with a repeated incidence of suspect home claims is more likely to make a suspect vehicle claim or travel insurance claim. At a larger scale a number of fraudulent vehicle insurance claims have identified large fraudulent commercial insurance claims.

3. Capture intelligence and learn.
All investigations provide invaluable intelligence, it is important to capture every outcome or interaction with your customer, whether a claim is denied, marked for additional attention or cleared. This insight is added to the networks and allows the system to learn and adapt scoring to improve accuracy. A networked database also allows third-party fraud flags to add far more value as they can be used to enhance risk scores even if they are not directly matched with the claim in question.

4. Use automated scoring at every step.
Cost can be reduced if you can confidently identify at the earliest possible opportunity which claims need further investigation and which should be paid. The system should re-score the claim every time any additional information is obtained throughout the claims handling process. This ensures that time is not wasted in unnecessary investigation and customers are not kept waiting for settlement.

5. Eliminate organized crime.
Insurers who implement networked analytics solutions find that organized crime represents a disproportionally large volume of claims fraud. It becomes very apparent that what initially appear to be opportunistic frauds are frequently connected and part of organized crime. A networked analytics approach is the ideal way to identify and deal with the conscious, pre-meditated activities associated with organized fraud.

6. Leverage free text documents.
Many insurers fail to make use of free text associated with claims, policies, customer interactions or contracts. This text can be mined and used in multiple ways for example, words and phrases can be used in risk scoring to better predict fraud and entities can be extracted from the text automatically to enhance networked data. Investigators can search huge volumes of text documents to find evidence, historic behavior or commonalities in seconds. Text analysis, categorization and heat mapping by claim types and geography can help to identify fraud or risk hot spots. The ability to process and integrate text is especially important in areas such as commercial, medical and liability related fraud.

7. Identify and remove supplier fraud.
Suppliers represent a significant share of claims spend, from garages, to doctors, lawyers and accident management companies. In many cases these are enablers of organized fraud, but they can also be fraudulent in their own right. By networking all of the claims around these entities and then performing a range of automated analytics such as outlier analysis, consistently anomalous behavior and identification of links to known frauds the identification of, what is frequently, large-scale fraud within the supplier base can be uncovered.

8. Prevent fraud from re-entering the system.
Fraudsters are very persistent and once they have been able to successfully deceive an insurer they will continue their attacks. By capturing denied claims or suspects onto the networks of customer data it is possible to run a match at policy inception. This can flag any connection of a new customer to a previous fraudster or ring of organized crimes and queue that particular application for exceptional checks before accepting the application. The same approach can be used with third-party known frauds or sanction lists to prevent money laundering.

9. Spot fraudulent brokers and insiders.
Sophisticated fraud rings will often work through brokers or place insiders within insurance operations to perpetrate volume fraud. For example, by adding the user ID's of the staff members who processed any part of the policy, to the networks of customer data, it is possible to identify any staff who are showing strange behaviors or who are overly connected to fraudulent activities.

10. Sanctions list checking.
To protect from potential reputational damage and financial penalties from the regulators, insurers have to ensure they operate an effective and auditable sanctions list checking process. By implementing a solution which provides a 'single view of customer' it is possible to effectively confirm an individual's identity and check them against any non-trade lists and reduce the burden of effective regulatory compliance.

Further Leverage
The depth of insight that is obtained by linking data across lines of business and through history to produce complete single views of customer or supplier can be used for other purposes too. Policy pricing, for example, could be dynamically adjusted based on risks to improve the quality of the customer base and marketing campaigns can be optimized through the analysis of cross product adoption or identification of 'influencers.'