Call Center Co-Sourcing: A JointPricing andStaffing Framework
Jeffrey P. Kharoufeh,Ph.D.
AssociateProfessor
DepartmentofIndustrialEngineering
UniversityofPittsburgh
Abstract
Increasingly, service organizations are electing to co-sourcesomeof their customer support functions, especially those handled bycallcenters.That is, ratherthanservicing requestsexclusively within-houseagents,aportion of service capacitycanbedelegated toanoutside service provider (or contractor) to reducecosts.Indeed, thebusiness of co-sourcingcustomerservicecentershasrapidlygrown into a multinational,multibilliondollar industry. However, organizationsmust weigh the economic benefits of co-sourcingagainsttheperceived(or real) costsofsurrendering control of their primarysource of direct customer support. Specifically, theymustdecide how much oftheoperationshouldbe co-sourced,andhow much should be kept in-house, when facedwith uncertainand dynamic demand.
Inthis talk, I will present a joint queueing- andgame-theoreticframework for analyzing the behavior of anexternalcontractor who serves multiple, independent call centers. The contractor seeksto setper unit agent holding and activation prices so as to maximizeexpected revenues over a finite contractperiod.Thecallcenters,facedwith stochastic arrivals, forecasting uncertainty and quality-of-service (QoS) requirements,respond to the contractor’s pricing strategy by setting staffing levels.Before the contract period begins, thecall centers decidehow many in-house agents to staff and how many co-sourced agentsto place on holdfor each period.Subsequently,theyreactivelyactivate agents during each period in order to satisfy theirQoS constraints. This problem is formulated as a Stackelberggame in which the contractor plays the part of the leader,and the multiple, independentcall centersrepresent the follower.It is shown thatthebi-levelprogramcanbereformulated so that theoptimalholdingand activation prices are obtained by solving a quadraticallyconstrained linear program. Thecallcenters’optimal staffing decisions are shown to be highly tractable underan expected delay constraint. Computational results illustrate the usefulness of the model, even in theabsence of perfect,symmetricinformation.
Biography
JeffreyP.Kharoufeh is anAssociate Professor in the Department of Industrial Engineering at the University of Pittsburgh. Hismethodologicalareasofexpertiseare applied probability and stochastic modeling with applications inqueueing theory, reliability theory,maintenanceoptimizationand renewable energy systems. Hisresearch has been funded by the National Science Foundation, the AirForce Office of Scientific Research, the National Reconnaissance Office, the Department of Veterans Affairsandother federal agencies.Dr. KharoufehearnedaPh.D.inIndustrialEngineeringand Operations Research at thePennsylvania State University where he was an inauguralWeissGraduate Fellow.He currently servesasArea Editorof Operations Research LettersandAssociateEditorofOperations Research (Stochastic Models). HeisaSenior MemberofIIEandaprofessionalmemberof the INFORMSApplied Probability Society.