Keller and Ozment (1999) Problems of driver turnover Costs $3,000 to $12,000 per driver Shipper effect SCM impact Tested solutions Pay raise Regional routes (swapping) Newer equipment Rewards for long stay
Study hypotheses Voice sensitive Exit sensitive Responsiveness Turnover
Voice Exit Turnover Responsiveness
Data collection Large TL carrier Pretest Top 100 US carriers 149 usable data
Voice Exit Turnover Responsiveness Results
Study Implications Significant impact of dispatcher on turnover rate High sensitivity to complaints and exits, and responsiveness lead to low turnover rate Train dispatcher for responsiveness Assign assistants to dispatchers (n > 50) Use inputs from exiting drivers
Questions 1.Why drivers quit-and-hire within the industry? 2.What are the costs of losing drivers for carriers? 3.If you are the management of a trucking company, what would you do to prevent or reduce driver turns? 4.How do you train dispatchers? What is your strategy for hiring new dispatchers? 5.What other factors should be considered when analyzing driver turns? 6. How does this study change the way you play simulation game?
Min and Lambert (2002) Driver turnover impacts Higher rate Newer equipment $ 446 billion industry 3.1 million drivers Study questions Data Randomly selected 3000 carriers – 422 responses Results
Questions 1.What kind of drivers do you want to hire or not want to hire? 2.How does the driver turnover affect the whole supply chain? 3.As the management, what would you do to prevent driver turns? 4.Would giving high pays to drivers solve the problem? 5.What other factors would you consider?
Predicting Truck Driver Turnover Suzuki, Crum, and Pautsch (2009)
Introduction Truck driver turnover is a key industry problem (TL). Many studies have investigated driver turnover. Limitations of past studies: (1) Static analyses (2) Survey data Missing an approach that: (1) uses time-series approach (2) utilizes operational work variables (data)
Advantages of using new approach (1) Operational work data = “revealed” data. (2) Data collection advantage. (3) Can assess dynamic effect of predictor variables. (4) Can be used as a practical decision tool. For these reasons several TL carriers expressed interest in providing data for analyses This paper reports results of two case studies and examine the effectiveness of this new approach from a variety of perspectives.
Questions to be answered (1) Are Operational work variables good predictors? (2) How do they compare against demographic variables? (3) Can the model be used as a practical decision tool?
Background (Carrier B) One of the largest TL carrier in the US. 150% driver turnover rate Tested almost all possible solutions Want to develop a method to predict driver exit for each individual driver by time Data mining method What else? ISU approach Application of the survival analysis (duration model) Predicts death (e.g., life expectancy) Time-series approach Quit prediction based on statistical probability
Data Weekly observations of all drivers (> 5,000) One-year data (52 weeks) Both stationary and non-stationary variables included Total sample = 117,874 Computation time = approx. 60 min (1.8 Ghz Pentium 4 PC).
Background (Carrier A) Medium TL carrier, with approx. 700 drivers. 80% driver turnover rate Wants ISU team to analyze their data and come up with recommendations for reducing driver turns. ISU Model Same model as that used for the large TL carrier. Good opportunity for ISU team to (1) examine the robustness of the previous estimation results, and (2) test the validity of the approach.
Data Weekly observations of all drivers (9 months). Both stationary and non-stationary included. Slightly different set of predictor variables Total sample size = approx. 29,000.
Implications Pay effect Dispatcher effect. Operational data effect Personal characteristic effect. Hire source effect Other noticeable effects? Demographic vs. Operational data
Model Validation Face validity Estimation robustness Macro-level validity Micro-level validity External Validity
Actions & Results (Carrier A) The carrier has changed its practice by using study results Action 1: Driver referral team Action 2: Incentive program for dispatchers Action 3: Improved information to dispatchers The turnover rate has improved. Actions & Results (Carrier B) Outperformed data mining method The carrier has implemented the ISU model. Seeking to combine the model with load-assignment model
Questions 1.How would you utilize the proposed driver-exit forecasting model to improve your turnover rate? 2.Does this type of model give benefits not only to each carrier but also to the whole industry? 3.What conclusions and implications can you drive from the two set of studies? 4.IS this type of model more helpful for large carriers than for small carriers? 5.What other factors would you consider in future studies?
Suzuki (2007) Introduction Driver turnover rate is still high and increasing. Many studies on this topic, but focused on how to improve turnover rates. By how much should the rates be reduced? “What level of turnover rate should carriers attain to generate desirable business results?” Develop a method of calculating a “desirable” or “target” turnover rates for motor carriers. Model Calculates the desirable rate for each individual carrier by considering the carrier’s unique characteristics. Based on statistical confidence (95%)..
Suzuki (2007) (1) (2) (3) RC = driver replacement cost M = net profit per day per driver = profit desired from each driver before exit = target operating profit margin RPD = revenue per driver per day
Suzuki (2007) Excel file with VBA Driver heterogeneity Tested the validity of the model for carriers with heterogeneous drivers. Results look promising (Table 3). Is your company’s turnover rate higher/lower than it should be?