Fraud Detection

The Significant Impact of Machine Learning In Warranty Fraud Detection

Daniele Micci-Barreca

The opportunity for Machine Learning in detecting warranty claims fraud.

Machine learning (ML) is a branch of artificial intelligence that looks at patterns of data and draws conclusions. Once it gets good at drawing correct conclusions, it applies itself to new data sets to find out hidden patterns. It is not a single technique or technology, but rather a field of science that incorporates numerous technologies to create systems that can learn from the data in their environment and then make predictions and take actions when confronted with a new situation.
How Machine Learning works
ML techniques can be categorized as supervised or unsupervised. Supervised algorithms require an analyst  to provide both input and desired output, in addition to furnishing feedback about the accuracy of predictions during algorithm training. The Data scientist determine which variables, or features, the model should analyze and use to develop predictions. Once training is complete, the algorithm will apply what was learned to new data that is being acquired from daily operations.

Unsupervised learning does not need to be trained with desired outcome data. It is used against data that has no historical labels. The system is not told the “right answer.” The algorithm must figure out what is being shown. The goal is to explore the data and find some structure within. Unsupervised learning works well on transactional data. For example, it can identify warranty claims and create profiles of typical claims for certain type of repairs.
Warranty Fraud situation today:
Warranty management has increased in importance dramatically over the years and the general maturity of the warranty profession in the industry has improved during the past 15 years. However, the battle against warranty fraud continues to be a challenge for many warranty professionals.

Industry studies show up to 10% of warranty costs are related to warranty claims fraud, costing manufacturers billions of dollars.

Most companies suspect warranty fraud, but are not sure of the extent and ways to eliminate it. The existing tools and methods to detect warranty fraud are rules based which allows you to identify only known fraud strategies or are complex and expensive, causing manufacturers to unwillingly bear the excessive costs.

To add to the equation, the sheer volume of claims makes it very difficult for knowledgeable warranty processors to thoroughly review and analyze individual claims.

Use of business rules during the claims process can identify some errors in claims entry, but are ineffective in mining the warranty data to detect anomalies and patterns that indicate fraud. Manufacturers need a powerful, easy-to-use, and cost effective warranty fraud detection solution.

Today fraudulent warranty claims occupy an estimated 3% to 15% of the average company’s warranty costs, which generally average between 1% and 4% of product sales. In individual companies, these figures can be much higher. Even at the low end of the range this translates to several billion US dollars globally making both warranty and warranty fraud major issues.
Case to use Machine Learning for Warranty Fraud detection
Businesses must improve the accuracy and speed of their decisions on fraudulent threats. The only mature technology available to achieve this is Machine Learning. Coupled with the fact that ML allows organizations to identify new fraud strategies that are being adopted, ML provides the perfect spoiler to organized fraud schemes in warrant claims.

In fact, one of the most effective uses of ML in identifying warranty fraud is staying ahead of new fraud schemes that the organization has not witnessed before.

Manual review is most effective as the last defense against fraud: While it can be invaluable, particularly in cases where there is no substitute for human insight, manual review works best to help fine-tune machine learning models decisions and aid their detection of changing patterns of fraud, rather than being the sole fraud detection process.

The feedback loop allows information from detected fraud threats to be used to enhance the detection engine so that instances of fraud of this type are identified much quicker and efficiently.

There is little doubt that where large amounts of data are concerned, machines are far more accurate and effective. They are able to detect and recognize thousands of patterns fraud schemes instead of the few captured by creating rules.. This is the reason why we use machine learning algorithms for preventing fraud for our clients. The three factors which explain the importance of machine learning are –

  1. Reacting at the speed of business – The velocity of commerce is rapidly increasing and will continue to do so in the foreseeable future, it’s very important to have a quicker solution to detect fraud. Our merchants want results fast so that they can act fast. It is so much easier to withhold payment till the issue is resolved than to pay and try to retrieve the payment after the fraud has been identified. Only machine learning techniques enable us to achieve that with the sort of confidence level needed to approve or decline a transaction.
  2. Handling increasing volumes of information – As is true in almost any field, the problem of identification does not lie in the extreme cases because they are easy to spot. The problem lies in cases that seem to be ‘on the fence’ – which is probably the bulk of your transactions. Machine learning algorithms and models become more effective with increasing data sets. Machine-learning improves with more data because the ML model can pick out the differences and similarities between multiple behaviors. Once told which transactions are genuine and which are fraudulent, the systems can work through them and begin to pick out those which fit either bucket. These can also predict them in the future when dealing with fresh transactions.
  3. Efficiency – Unlike humans, machines can perform repetitive tasks with the same degree of efficiency and accuracy throughout. Similarly, ML algorithms do the dirty work of data analysis and only escalate decisions to humans when their input adds insights. ML can often be more effective than humans at detecting subtle or non-intuitive patterns to help identify fraudulent transactions. Moreover, unsupervised ML models can continuously analyze and process new data and then autonomously update its models to reflect the latest trends.

The ideal combination – the smart assistant
Machine learning is not a panacea for fraud detection. It is a very useful technology which allows us to find patterns of an anomaly in everyday transactions. They are indeed superior to human review and rule-based methods which were employed by organizations.

The ideal situation is for the ML based fraud detection solution to be a feeder to the experienced human eye. The ML solution would throw up ‘suspicious’ warranty transactions for human review. The expert human eye would make the judgement on whether or not the transaction is a fraud thereby bringing their knowledge to bear on the process but at the same time ‘teaching’ the ML solution to further refine their selection criteria to detect fraud more efficiently in the future.

Daniele Micci-Barreca