Policies for Customer-Preference-Based, Stochastic Scheduling of Field Visits



Motivated by the sales force operations of a B2B payment services company in Belgium, we develop a model for scheduling appointments for visits for a team of sales agents. The appointments are made by an outbound call center and subject to uncertain customer approval. The objective of the model is to maximize the number of appointments per agent by minimizing the travel and idle times between appointments. We formulate the model as a Markov decision process, demonstrate the existence of an optimal policy, develop an upper bound on the optimal performance, and analytically derive sensitivity properties with regard to several design parameters such as the length of the scheduling horizon, the capacity of the call center, and the replenishment quantity of the client database. Because of the large amount of continually incoming data, the problem suffers from the curse of dimensionality, even for small instances. Therefore, we develop two time-efficient scheduling policies that dynamically decide which clients to call and which time slots to propose to each client. The two policies differ in their computational complexity and performance. We perform numerical experiments based on a real data set to test the performance of the policies. We find that the appointment scheduling policy is an effective way to achieve a balance between effectiveness, efficiency, and reactivity of the field force; a balance that can be tailored to the operations strategy of the company.