Mark Grabau’s scientific contributions

Many insurers pay independent agencies a contingent commission based on the agency's performance. The insurer must accrue funds during the year to recognize the expected contingent commission payout in the financial statements. Westfield Insurance wanted to reduce their accrual forecasting error from 20 percent to less than five percent. We built a Monte Carlo simulation to simulate each Westfield agency's performance. We clustered agencies into representative groups as a proxy for generating correlated random variables and then designed an experiment to shift statistical distributions of agency key performance indicators. The results of the simulation were then used to derive a formula for forecasting expected contingent commissions based on overall company results. The approach, results, and areas for further research are discussed.

Westfield Insurance operates a call center that handles billing and claims calls. The call center has resources dedicated to each call type and hybrid resources that can handle both call types. The use of hybrid resources makes it challenging to predict what impact a staffing change will have on service level metrics. This paper documents the approach, lessons learned, and business impacts of modeling the call center with a discrete event simulation. It also provides a method for using queues and resource sets to model hybrid resources.

Westfield Insurance is undertaking a replacement of its legacy claims system. They had several business process and staffing changes they were considering; however, they had no way to test their options prior to implementation. After successfully simulating the First Notification of Loss (FNOL) to adjuster assignment process, Westfield decided to append the rest of the claims process. This resulted in an end-to-end claims process simulation, including adjudication, settlement, litigation, salvage, subrogation claims and fraud. The process, results, and lessons learned are discussed.

Scheduling games for any professional sport is an extremely complex process. Scheduling recreational softball games for over 1,000 players is also a difficult process. A scheduler must adhere to nearly two dozen intertwined scheduling rules, while simultaneously ensuring that the players play their allotment of games. This paper reviews the scheduling of college basketball and professional baseball, football, and soccer. It then presents an integer programming approach to solve this problem. Finally, it identifies the obstacles encountered and the qualitative and quantitative benefits of the approach.