Analytical Tools For Auction Decisions
Business Background:
Auctions have been used as market mechanisms for selling/buying a wide variety of goods and services as well as for allocating contracts. With the recent explosion of electronic markets, there has been a growing interest in auctions as convenient mechanisms for buying and selling goods and services over the Internet. Many enterprises are planning to conduct a substantial percentage of direct and indirect procurement operations through online auctions. Auctions are also used in disposal of excess inventory by an increasing number of companies and they account for increasingly larger share of total dollar volume. Thus, a growing number of firms need to make decisions on when/whether to use auctions and how to use them. Optimal configuration of auction parameters is of important business interest.
Problems Addressed:
To conduct an auction one has to make many decisions - one has to select an auction format, a reserve price, bid increment/decrement, to name a few. In multiple item auctions the decision problems increase substantially: Should one sell/buy multiple lots sequentially or all at once? If sequentially, in what order? Should one bundle multiple items into lots? If so, how?
The key challenge is that in auction design 'one size rarely, if at all, fits all.' The best choice for any of the decision problems depends on the market environment - the number of bidders, the nature of uncertainty and private information held by the bidders, the bidders' attitudes toward risk etc. Furthermore, depending on the market environment, the effect of wrong design choices can be substantial.
Large portions of this decision space are still uncharted territory - both from analytical and from practical business points of view. The goal of the Auction Design Toolbox project is to develop a set of analytical tools to process available information to guide the customization of the auction parameters to the specific needs of the user.
Our Contribution:
The methodology that underlies the Auction Design Toolbox draws on various knowledge sources ranging from game theory, economics, econometrics, mathematical programming, statistics, and computational and experimental methods.
Analytical tools for auction design:
- Techniques for identification and estimation of unobserved random elements of the auction environment from auction data
- Methods for elicitation of information from domain experts about the key structural elements of the auction environment
- A computational workbench to simulate a variety of common auction processes
- A library of bidding models for various auction mechanisms and auction environments
Applications in procurement auctions:
The analytical tools in Auction Design Toolbox have been used in various combinations to provide decision support in the following decision problems in sourcing and procurement auctions:
- Determination of optimal reserve prices
- Selection of the best auction format
- Optimal number of bidders
- Optimal sequencing of multiple lot auctions
- Selection of feedback rules in open ascending/descending auctions
- Determination of optimal lot size and lot composition for multiple item auctions
- Winner determination in multi-item multi-criteria auctions
Contacts:
Proj Mgr: Fereydoon Safai
Project Lead: Kemal Guler
|
printable version
