Last Modified
  • March 19, 2019
  • Pendem, Pradeep
    • Affiliation: Kenan-Flagler Business School
  • In 2015, on a global level, the service industry represented 66% of GDP and employed over 51% of the total working population making it one of the largest industries. Some of the sectors in this industry include finance, transportation, call centers, retail, and health care. Customers and service providers are key players in this industry. A successful transaction between these two results in a valued service for the customer and revenue to the provider. The primary objective of the providers, therefore, is to understand the customer’s needs, meet their requirements, provide quality service and achieve customer satisfaction. In this thesis, I utilize large data sets on customer-provider transactions to study two important issues. First, I build micro-demand models to predict true demand incorporating customer behavior, time and spatial dynamics. I utilize the predicted demand to optimally allocate the resources for improved operational performance. The study contexts I focus on are Bike Sharing systems and Street-Hail Taxi services. Second, I build micro models to understand the factors driving customer provided satisfaction measure on logistics service and their impact on purchase probability in E-commerce platforms. In the first chapter, I analyze the optimal allocation of bikes in a network of stations to improve ridership under non-stationarity demand and station substitution. Using large datasets on the censored trip and minute-level inventory, walking distance between stations and a stochastic model, I predict true demand at each station. Then I determine an optimal allocation of bikes across stations at the start of the day utilizing a dynamic program to maximize ridership in the network. I find the optimal policy could improve ridership and service level by 7.60% and 1.69% respectively. In the second chapter, I examine the impact of logistics performance metrics such as delivery delays, customer's promised speed of delivery, order split, etc. on logistics service ratings of sellers on an e-commerce platform. Using a large dataset of customer orders from an e-commerce platform, I find logistics ratings are negatively impacted by delivery delays, but positively impacted by faster-promised speed of delivery and total order amount paid. I also find that logistics ratings impact customer purchasing behavior positively. Lastly, I show that a reduction in delivery delay by one day can improve the average weekly sales by as much as 2.5%. In the third chapter, I study passenger demand estimation problem in Street-Hail Taxi services. I utilize large-scale datasets on GPS information of pick-ups and drop-off from New York Yellow Taxi services for this study. I first develop a stochastic model (double-ended queue) to predict passenger demand in location and time. The model allows for non-stationarity, randomness in arrivals and reneges of both drivers and passengers. Using sample path information along with Maximum Likelihood Estimation, I develop a framework to estimate true passenger demand, drivers and passengers renege rate. The predicted demand can be used to analyze the optimal timing of drivers change their shift to maximize revenue under the current status quo of delay in shift changeover. Overall, the thesis focuses on the analysis of large and granular transactional data to build micro demand models incorporating customer behavior and incorporate the models into planning to improve operational efficiency.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Kulkarni, Vidyadhar
  • Kesavan, Saravanan
  • Mersereau, Adam
  • Deshpande, Vinayak
  • Staats, Bradley
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2018

This work has no parents.