Table of Contents Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Objective. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Assumptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Model Development. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Input Data Collection and Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Jamie Serenson Page 1 of 8 Executive Summary Decreasing the automotive loan application processing time to 30 minutes would offer a large competitive advantage to First Security Bank. Discreet event simulation was used to simulate the current system. Once a model was validated, system modifications were evaluated in an attempt to decrease the loan application processing time to 30 minutes. The following recommendations were made to achieve the 30-minute loan processing time objective: Regions 4 and 1 should add one and three additional loan officers, respectively. Additional training should occur and individuals compensation incentives should be tied to: Reducing the mean data entry service time to 6 minutes with less than 5% errors. Reducing the mean regional loan office service time to 20 minutes. Introduction First Security Bank recently introduced a new consumer lending software package to increase the speed with which an auto loan application can be processed. The objective of the program was to reduce the application processing time to 30 minutes. This would increase revenue by facilitating sales at the car dealerships. Appendix 1 represents the current process. The loan applications enter the system via incoming fax machines. A runner then brings the application to a data entry clerk that puts the application in electronic format. The electronic application is then sent to the regional loan offices. Finally, the processed application exits the system via fax machines at the dealerships. Please refer to Appendix 1 for a description of the current capacity, processing times and goals for the current process. Jamie Serenson Page 2 of 8 Objective The objective of this project is to make recommendations for decreasing the average loan processing time to 30 minutes by utilizing discreet event simulation. Scope This project details a strategy for reducing the loan application processing time to 30 minutes. To accomplish that, the current system was modeled using Promodel software. The model includes all relevant steps that have an impact on application processing at First Security Bank. The model was designed to be as simple as possible while still representing all key aspects of the system. Once the baseline model was validated, the effects of process changes were evaluated on the loan processing time in an effort to reduce the loan processing time to 30 minutes. Requirements To perform this analysis it was necessary to be familiar with Promodel 4.2 and to have access to the software. The model created is small enough that the student version of Promodel can be used to run or modify the current model. Assumptions 1. The fax transmissions were given to be less than two minutes in length. It was assumed that the distribution of times as U(1.5, 0.5) for incoming and return faxes. A uniform distribution was chosen because the service time was very short and there was not a lot of information known concerning the distribution. 2. No service time distribution was listed for the runner. It was assumed that the runner did not impact the overall service time of the system. Jamie Serenson Page 3 of 8 3. The distribution of the service times for the data entry clerks was assumed to be normally distributed. It was known that the mean service time for the data entry clerks was initially 9.5 minutes and the standard deviation was assumed to be 1 minutes. This is a coefficient of variation of approximately 0.10, which was believed to be realistic. 4. The distribution of the service times for the loan processing office was assumed to be normally distributed with a is a coefficient of variation of 0.10. It was known that the overall loan processing time should be 37 minutes initially, so the actual service time for the loan processing office was solved for with trial and error. Model Development There are six different regional loan offices that have different staffing levels and processing times. Although it is possible to model the system with the six unique regions, it would simplify the model tremendously if the six regions could be consolidated into one multicapacity location that was representative of the actual system. This was, consequently, investigated before beginning to design the model. As can be seen in Appendix 2, the average processing time per regional office is approximately 37 minutes if the workload remains below 740 applications annually per loan officer. If the annual workload decreased below 740 application per year, the mean loan processing time remains constant at 37 minutes. Once the annual workload exceeds 740 applications, the mean processing time increases substantially (see Regions 4 and 1). Based on that, it was determined that the staffing levels at each regional loan office should be set such that each loan officer had an annual workload less than 740 applications. Appendix 3 illustrates that to accomplish that Regions 4 and 1 will need to add one and three additional loan officers, respectively. Jamie Serenson Page 4 of 8 By adding the additional four loan officers, the mean loan processing time will decrease 22% from 47.5 minutes to 37.0 minutes. With the new staffing levels at the regional loan offices, it will be assumed that all regional loan officers will have the same processing time so they will be modeled as one multi-capacity location. Input Data Collection and Analysis Due to my limited exposure to modeling before taking this class, a case study was used from “Simulation Using Promodel” (Harrell, p370). The input data was given in that problem. Appendix 1 presents the current capacity, processing times and goals for the loan application process. Table 1 describes the current workload distribution and effectiveness (average processing time) for the six regional dealerships. TABLE 1: CURRENT WORKLOAD DISTRIBUTION AND AVERAGE PROCESSING TIME PER REGION Region Applications Loan Officers Average Processing Time [min] 1 6,150 6 58.8 2 1,485 2 37.2 3 2,655 4 37.0 4 1,680 2 51.1 5 1,440 2 37.0 6 1,590 3 37.0 Total 15,000 19 Verification Based on the given information, a first draft of the model (Model 1) was created. The mean processing time per application was 33.0 minutes. It should be noted that all output files are for a 40 hour simulation with 10 replicates. A disk is also attached that has copies of the model and various other files. Jamie Serenson Page 5 of 8 To verify the model, first the model code was reviewed in detail to ensure that the commands were entered as intended. The animation was then reviewed during simulation to ensure that the entity flows in the system visually appeared to be correct. Finally, the output of the simulation was reviewed to ensure that the actual performance metrics were reasonably close to the estimate performance metrics. The model was verified to be working correctly. Validation To validate the model, the model was compared to the system. For the initial model created, the mean processing time was 4 minutes less than the expected value of 37 minutes; however, the process time required for the regional loan offices was not currently defined. To set the mean processing time to 37 minutes, the service time at the regional loan offices was set to N (24,2.4). Once the mean processing time for the system was consistent with the case, the model was considered to be validated. Results The key performance metrics and the total loan process time are presented in Table 2. TABLE 2: RESULTS FOR DISCREET EVENT SIMULATIONS Data Type Performance Metric 1 Input Output Percent Data Entry Errors Data Entry Service Time, min Loan Office Service Time, min Total Loan Process Time, min 2 Model 3 4 5 10 10 5 0 5 N(9.5,1) N(9.5,1) N(6,0.6) N(6,0.6) N(6,0.6) N(20,2.0) N(24,2.4) N(24,2.4) N(24,2.4) N(20,2.0) 33.0 37.1 33.2 33.0 29.2 Discussion Since the baseline model was established, the process output was reviewed in detail to identify areas that merited further investigation for improvement. It was observed that the Jamie Serenson Page 6 of 8 maximum contents of the incoming and outgoing fax machines were half of their current capacity. This low utilization was consistent with the goal of no busy signals (failed arrivals). Based on this, the number of incoming and return fax machines did not need to be modified to decrease the loan processing time or to meet the no failed arrivals goal. The utilization of the data entry clerks and regional loan officers was very low at 10-15% (see Figure 1). This indicates that adding more data entry or loan officer personnel will probably not decrease the loan application processing time. It is assumed that the data entry clerks and regional loan officers have other responsibilities that utilize the remainder of their time. It will is also assumed, however, that they have sufficient time to manage their current responsibilities and that adding additional data entry clerks or regional loan officers would not significantly impact the loan application service time. Figure 1: Location Utilization for FSB Coin To reduce the number of errors from 10 to 5% and the mean processing time from 9.5 to 6 minutes, it is recommended that First Security Bank provide additional training for the data entry clerks and formally tie their compensation incentives to achieving these goals. If these Jamie Serenson Page 7 of 8 objectives can be met the mean loan application processing time will be reduced to 33.2 minutes. Based on the current performance, it is unlikely that the data entry clerks can decrease their processing time below 6 minutes and bring errors below 5% without a major systems change. Even if they could, it would not have a large impact on the mean loan processing time due to the relatively low occurrence of errors. Completely eliminating the errors would only decrease the mean processing time to 33.0 minutes. Since the regional loan offices represent the majority of the processing time and they are failing to meet their target service time, it would be also required to investigate the processes at the region loan office to reduce the mean service time to 20 minutes. If the mean service time at the regional loan offices is reduced to N(20, 2), the 95% confidence interval for the mean processing time of Model 7 is 29.1 to 29.4 minutes. Conclusions The data entry clerks are not meeting their target service time. The regional loan offices are not meeting their target service time. The staffing level at each regional loan office should be set such that each loan officer had an annual workload less than 740 applications. Recommendations Regions 4 and 1 should add one and three additional loan officers, respectively. Additional training should occur and individuals compensation incentives should be tied to: Reducing the mean data entry service time to 6 minutes with less than 5% errors. Reducing the mean regional loan office service time to 20 minutes. Jamie Serenson Page 8 of 8 References 1. Harrell, Ghosh and Bowden. “Simulation Using Promodel”. McGraw Hill Publishing Company. Copyrighted 2000. Page 370-371.