Back in the Driver’s Seat
Auto manufacturer saves millions in warranty claims using predictive analyticsDOWNLOAD
Insurance fraud is on the rise in many industries – and automobile manufacturing is no exception. Fraudulent warranty claims can represent a significant cost to businesses and industry specialists estimate that inappropriate claims represent three to five percent of all submissions1. To most auto manufacturers, improving the claim adjudication process can mean tens of millions of dollars in savings every year – savings that can go directly to the bottom line.
After sifting through one million warranty claims per month from 4,600 dealerships, totaling $1.8 billion in just one year, this manufacturing client realized that its existing process wasn’t working. It had previously based its claim adjudications on long-term experience and internal expertise, but believed that there was a significant opportunity to improve operations through more sophisticated analysis.
This auto manufacturer came to Deloitte with its challenge and we determined that its warranty claims pattern detection process could be greatly improved by using advanced analytics – statistical, mathematical and other algorithmic techniques – to examine the ways in which specific business issues relate to data on past, present and projected future actions. Predictive modeling tools and processes would also provide a measurable improvement over the manufacturer’s current claims adjudication process, which relied heavily on manual reviews and static rules.
The large number of parts, labor categories and things that can go wrong in a vehicle present manufacturers with an unending combination of ways to potentially fall victim to deception. The company’s existing system of data mining and human analysis simply wasn’t robust enough to identify, flag and reject a high quantity of questionable claims. Additionally, the increasing volume and technological advancements in warranty processes (multiple car models, online submissions, etc.) make it even more difficult to identify inflated, erroneous, or inappropriate claims in a timely and economically efficient manner.
With this in mind, Deloitte was challenged to go beyond the few thousand rules already in place when searching for an oddity or anomaly in the claims data and to find new rules that would become part of a new predictive modeling system. As shown by successes in other industries like retail and transportation, advanced predictive modeling techniques can transform large amounts of data into actionable insights for management. We looked to apply these same techniques to the warranty business to quickly generate results that would be nearly impossible to obtain through traditional analysis.
Solving such complex problems would take not only deep advanced analytics tools, skills and experience, but also a profound understanding of the auto industry and its warranty processes. The project team drew upon the depth and breadth of the firm by seeking knowledge from its industry and analytics specialists. Guided by this strong industry experience, the team applied algorithms used in other industries, such as retail, telecom and financial markets, to the auto warranty business.
Deloitte also used advanced analytics tools and techniques to analyze several years of historical claims data and design warranty verification rules. We reviewed more than 16 million claims using advanced modeling techniques such as classification and regression trees, association rules algorithms and non-linear regressions and used fictitious populations of numbers in order to develop new rules to recognize false or inappropriate claims. The result was substantially reduced payment of false or inappropriate warranty claims. However, developing the algorithms was only part of the solution.
In addition to the predictive models, it was necessary to develop the documentation containing the detail and logic for the new rules so they could be implemented promptly within the company’s systems. Additionally, a benefit analysis, including potential cash savings and the complexity of implementation, gave management the tools needed to help prioritize its actions and optimize internal resources.
This seamlessly integrated project approach allowed for innovative solutions that generated millions of dollars in savings for the client.
By using predictive analytic technology applications, this company improved its ability to identify potentially fraudulent claims, determine which claims required additional review and refine claim and related business processes to handle those transactions that are marked for review.
Working with the client, we created five new groups of rules, some of which contained several hundred rules that could be applied to incoming data. These new sets of rules will potentially generate more than $6 million in annual savings – more than twice what was initially expected by the company. In fact, our client believes that the final opportunity will surpass $10 million in annual savings.
In addition, we produced a pre-approved list of operations for each repair that is now available as a checklist for claims personnel and for the dealerships who submit these warranty claims. This will further support the claims process and help reduce the number of fraudulent claims.
The success of this project demonstrated that similar analytic processes can be applied from one industry to the next. When there’s too much data, companies cannot rely simply on human analysis. Spending time and energy developing the right systems will not only help save money, but can also put the organization back in the driver’s seat.
1 Westphal, Christopher. Data mining for intelligence, fraud & criminal detection: Advanced analytics & information sharing techniques. (2008). Boca Raton, FL: CRC Press, Inc.
As used in this document, "Deloitte" means Deloitte Consulting LLP, a subsidiary of Deloitte LLP. Please see www.deloitte.com/us/about for a detailed description of the legal structure of Deloitte LLP and its subsidiaries.