This is the second of the seven parts of blog post series “A practical guide to tackle auto insurance fraud”.
While Data Management and Data Quality are the basis for every analytical journey, and this becomes even more true for fraud detection analytics, the domain knowledge and business expertise plaid always a very crucial role for claim handlers and fraud investigators. This 2nd article of the series analyses the basic concepts, analytics methodology and techniques, for tackling the “known” fraud, the fraud typologies and fraud indicators that traditionally utilized in the battle against fraud.
Business Rules basic concepts
For a common understanding, let's first define what a Business Rule is (source: Wikipedia).
“A business rule is a rule that defines or constrains some aspect of business and always resolves to either true or false… Business rules can apply to people, processes, corporate behavior and computing systems in an organization, and are put in place to help the organization achieve its goals.”
In the insurance fraud area, a business rule consists of one or more fraud indicators that based on practical experience and observation, trigger suspicious for possible fraud behavior.
An example of a simple business rule can be a submission of a claim by a new customer, within 15 days upon the policy activation date. Based on field experience, such claim have increased probability to be fraudulent (the damage or accident maybe existed prior, not paid by previous insurer and now the claimant tries to re-bound and get paid back).
This rule consist of two data elements and two calculation steps:
- the date of claim’s announcement, the effective policy start date and
- the difference of two dates X and the comparison of X<15. If X<=15 then this business rule scores 1, otherwise 0.
With this approach we can build or utilize business rules libraries and categorize them in: data areas, severity segment, complexity of calculation and risk score. With the use of analytics and having already covered the fundamental step 1 (data management and data quality covered in the previous series article), these rules can be coded in specialized software and cross-check their validity against each transaction, either in a claims application for claims fraud or in a policy application for underwriting fraud or other. Further on this analytics process can be automated, can be executed either in a batch mode by night or even in real time.
In an upcoming article we will drill down to the rule fraud risk scoring analytic technique and how an insurer can calculate a unique score per rule, taking into account his own claims portfolio and business environment.
Business Rules areas
Fraud detection business rules have to be categorized in areas and have to be supported by source data. Examples of common rules areas are:
- Claim data related (relevant to the accident time, date, announcement date, accident scene characteristics, bodily injury or not, recent change in coverage, high value, theft, fire, suspicious location etc.).
- Vehicle data related (total loss, commercial etc.)
- Policy related (X accidents in the past months etc.)
- Supplier / body shop
- Individuals (customer, driver, claimant, participants, lawyer, agent etc.)
- Entities (telephone numbers, addresses, tax id, VIN etc.)
- Network (links between entities, individuals, claims, policies etc.)
However it is important the business rules that the insurer will select and apply, to be supported by available source data in good quality and make business meaning for his specific portfolio and environment. An insurer should not focus in adopting ‘any’ business rule and produce a, exhausting large set. This will need extensive effort and time to be coded and implemented, but also will trigger significant complexity in fraud risk scoring and alert generation. An insurer will have to utilize the minimum set of rules that will produce tangible results and accurate fraud alerts. Gradually will build upon it and expand the analytics system with new rules, as fraud detection is a constant battle and an insurer has to adopt its tactics constantly.
Another technique for “known” fraud is Watch Lists, meaning lists of individuals or entities that are proven fraudsters or have high probability to be fraudsters, based on historical own data or external data or other information. Each claim data have to be cross checked with a series of watch lists in order to identify if there is any link which triggers increased suspicious for new fraudulent activities.
Again there are several types and areas of watch lists, e.g.: individuals, suppliers/ body shops, lawyers, agents, vehicles etc. Each list can consist of several data elements, like: name, telephone number, address, tax id and other.
The analytics techniques and specialized data quality software that have to be utilized for comparison of claims data with the lists, are similar with the Data cleansing and matching techniques analyzed in 1st article of the series.
Tackling known fraud with business rules analytics and watch lists is a mandatory step for every insurer in a fraud analytics journey. Specialized fraud prevention software include such rules libraries, with concentrated domain knowledge from many years of R&D and field experience. However, each insurer is a different kind of story, so specific customizations and calibrations have to be applied. Fraud is a constant battle, so the best tactic is to start with a proper, not exhausting, set of rules and then build gradually upon that.
This is the 2nd post in a 7-post series, “A practical guide to tackle auto insurance fraud”. This series explores 7 analytics best practices techniques that insurers need to follow for tackling auto insurance claims fraud. Next post deals with the “unknown” fraud and the Advanced Analytics Techniques, for uncovering hidden fraud patterns and outliers.
Don’t miss to watch the on-demand Insurance Fraud webinar series that were completed on September.