The Chief Operating Officer (COO) of the Life Insurance business was livid and blurted “How the heck did this village bumpkin crack my system to do this fraud. This can’t go on. We will be ruined and our reputation will be lost”. The complexity of the fraud really stared at him. “We need to find a solution to prevent this and also support the business to grow exponentially”.
The Group Head Risk, could empathise with him and he made a bold statement “Let my crack team give you a solution that will give you peace and let you focus on the business”. The confidence and energy was suddenly high and the challenge appeared conquerable.
Being in retail business, the private Life Insurance service provider operates in a very competitive environment and frauds eat into the profitability.
Customers from all walks of life can take insurance cover and have different requirements, intentions and behaviour. It is challenging to manually understand and identify fraudulent intentions (aka greed) of a large range of customers.
COO team observed many patterns in historical fraud cases ranging from increased propensity of fraud during month-end pressure, to the role of collusion among distributor agency and involvement of doctors.
A multifunctional team with domain experience and knowledge was formed to create an innovation solution for identification of fraud and provide early warning signals.
Machine learning tools and artificial intelligence tools along with Statistical best practices were combined powerfully to create the innovative solution that had scalability. A two pronged attack of analysis of past cases and emerging fraud patterns through outlier detection techniques was rolled out.
Once a fraudulent pattern has emerged in a particular region the model could learn and predict similar trends in other regions and immediately raise a red flag for the same. Several ratios were identified that were used to identify potential red flags of a fraudster. The Machine learning could identify several such behaviour patterns that now could be recognised as suspects for further investigation. The model helped us in identifying the wolf in sheep’s clothing i.e. could sieve a fraudster from genuine customers.
By adopting the innovation model we could sieve the bad customers from good one, thereby reducing the cost of poor quality (prevention cost, failure cost). It also improved the TAT for processing, as good customers were allowed straight through approvals and thereby greater customer satisfaction and reputation.
The COO called the Group Head and said “Thank you for making the fraudsters Rest in Peace (RIP). You have really taken this monkey off my back”. There were smiles around and a sense of readiness for what could lie ahead.
Lessons Learned
- Innovation is necessary to stay ahead in the game called business
- All that glitters is not gold. One needs to really understand customer profile, behaviour and motives and this can be augmented through machine learning capabilities
- Greed has no limits and fraudsters will go to any extent to break your system
- Unless individual learning is converted to institutionalised learning the system will falter under pressure
- Innovation is easily possible when learning from multiple domains are applied seamlessly
- A cross functional team working together with a common purpose is far likely to succeed than working in silos
- The dichotomy of technology as a destroyer and an enabler will be more acute in the future
- Good customers pay the price for deeds of the bad customers. Hence one needs to weed out bad customers