Price and service discrimination in queueing systems: incentive compatibility of Gcμ scheduling

Research output: Contribution to journalArticlepeer-review

91 Scopus citations


This article studies the optimal prices and service quality grades that a queueing system - the 'firm' - provides to heterogeneous, utility-maximizing customers who measure quality by their experienced delay distributions. Results are threefold: First, delay cost curves are introduced that allow for a flexible description of a customer's quality sensitivity. Second, a comprehensive executable approach is proposed that analytically specifies scheduling, delay distributions and prices for arbitrary delay sensitivity curves. The tractability of this approach derives from porting heavy-traffic Brownian results into the economic analysis. The generalized cμ (Gcμ) scheduling rule that emerges is dynamic so that, in general, service grades need not correspond to a static priority ranking. A benchmarking example investigates the value of differentiated service. Third, the notions of grade and rate incentive compatibility (IC) are introduced to study this system under asymmetric information and are established for Gcμ scheduling when service times are homogeneous and customers atomistic. Grade IC induces correct grade choice resulting in perfect service discrimination; rate IC additionally induces centralized-optimal rates. Dynamic Gcμ scheduling exhibits negative feedback that, together with time-dependent pricing, can also yield rate incentive compatibility with heterogeneous service times. Finally, multiplan pricing, which offers all customers a menu with a choice of multiple rate plans, is analyzed.

Original languageEnglish (US)
Pages (from-to)1249-1267
Number of pages19
JournalManagement Science
Issue number9
StatePublished - Sep 2000

ASJC Scopus subject areas

  • Strategy and Management
  • Management Science and Operations Research


Dive into the research topics of 'Price and service discrimination in queueing systems: incentive compatibility of Gcμ scheduling'. Together they form a unique fingerprint.

Cite this