Economics of Resource-Pooling in IT Applications and Services

Many IT applications and services are provisioned via resource allocation mechanisms based on sharing of resources among active tasks. Resources are mapped to requests for fractional time intervals, and the mapping is revised in real-time as dictated by instantaneous demand and supply. This fine-grained resource sharing is made possible by efficient algorithms for multitasking, multiplexing and dynamic resource allocation. Such allocations increase technical and system efficiency but have some economic limitations. They increase uncertainty about quality of service (QoS) which creates downward price pressure and makes it difficult to price discriminate or to segment customers based on differential QoS. Moreover, under flat-rate price structures (which are widely used in IT services), light users bear an unfair share of the price while heavy users cause a negative effect on profitability (because they impose greater costs on the system) and cannot be deterred simply by raising price.

Our research looks at product design and pricing mechanisms to mitigate the negative economic effects while maintaining the essential technical benefits of resource pooling. When customers are heterogeneous in their usage patterns, and heavy users impose substantially greater workload than light users, firms can use forced bundling to create unfavorable conditions for heavy users, inducing them to drop out of the market. Similarly, long-run statistical QoS guarantees can be used to segment the market in a way that heavy users pay a higher expected price than light users, even under a flat-rate price mechanism. Capacity planning under resource-pooled environments with demand uncertainty can be improved through multi-part tariff structures and incentives for information revelation.

• 
  
• When customers are heterogeneous in their usage patterns, and heavy users impose substantially greater workload than light users, firms can use forced bundling to create unfavorable conditions for heavy users, thereby making them self-select to drop out of the market. A good illustration of this strategy is AOL's bundling of dial-up Internet access with a proprietary non-standard connection manager: power users hated this software and consequently stayed away from AOL service.
  
  H. K. Bhargava and J. Feng (2005), “America Online's Internet Access Service: How to Deter Unwanted Customers,” Electronic Commerce Research and Applications, Vol. 4, No. 1, pp. 35--48.
  
  
• One negative consequence of resource pooling is greater uncertainty about QoS. In the IT industry one way to manage this uncertainty is to offer performance-based price contracts which specify a guarantee and a price rebate in case actual performance falls below the guaranteed level. This is a profitable strategy when the market underestimates the firm's performance. Performance-based pricing can be used to signal the firm's true quality to the market.
  
  H. K. Bhargava and S. Sundaresan, “30 Seconds or Free! Contingency Pricing for Information Goods and Services under Industry-wide Performance Standard,” Journal of Management Information Systems, Fall 2003.
  
  
  
• Performance-based pricing, linked to statistical guarantees about quality of service, can also be used to segment markets in a way that heavy users pay a higher expected price than light users, even under a flat-rate price mechanism. This mitigates another negative consequence of resource pooled services: the difficulty in offering differentiated services or prices.
  
  H. K. Bhargava and D. Sun, "Performance-based Pricing in Shared-Resource Service Environments: Fair, Efficient, and Profitable", Working paper, UC Davis, last revised March 2006.
  
  
• Resource pooling underlies the idea that business information systems may, in the near future, utilize a utility computing model where computational services are provisioned through a collection of resources made available on a computational grid. Price structures for such markets will need to manage substantial demand uncertainty, and balance risks associated with capacity planning and demand commitment while creating incentives for informed parties to share private information that leads to more efficient planning and allocations.
  
  H. K. Bhargava and A. Bagh, "Tariff Structures for Pricing Grid Computing Resources", Proceedings of The Third International Workshop on Grid Economics and Business Models (GECON 2006), Singapore May 16, 2006.
  
  H. K. Bhargava and S. Sundaresan, “Computing as Utility: Managing Availability, Commitment and Pricing through Contingent Bid Auctions,” Journal of Management Information Systems, Vol. 50, Issue 3, Fall 2004.