IBM algorithms boost B2B e-commerce

IBM Corp. Research has developed two powerful algorithms specifically designed for business-to-business e-commerce applications that allow vendors to conduct advanced auctions as part of their procurement processes.

Now in limited beta testing, the Combinational Auction and Volume Discount Auction algorithms allow users to quickly assess the best prices for multiple items being offered from multiple suppliers. They can then more easily mix and match the products they need from across all their suppliers' bids. IBM intends to incorporate the algorithms into its WebSphere Commerce Suite Marketplace Edition.

"Customers have different ways of buying things from their many suppliers and have different business rules for each. But the goal is always 'how much can I get at the minimum costs while satisfying their demands'? We think this technology helps greatly with that process," said Chae An, director of e-commerce at IBM Research in Yorktown Heights, New York.

With the Combinational Auctional algorithm, a buyer in the food processing industry needing to buy a series of ingredients at a single bundled price from one supplier, for example, can easily compare and contrast various mixes of ingredients from other suppliers to decipher the best deal.

"So when a buyer has to decide which combination of bids he accepts based on the minimum cost but that also satisfy his other business goals and constraints, he can run this optimization [algorithm] to see what he should accept," An said.

Another example might be that of an automaker that must procure US$50 million worth of materials a year. Using the algorithm, it would be reasonable to assume automakers could save 5 percent on purchases over a year, a significant savings in that or any other industry with thin margins.

The Volume Discount Auction algorithm specifically focuses on sifting through a variety of different volume discounts being offered by suppliers, helping users decide which one is the best deal given their particular goals and business constraints.

The algorithms took about five months to develop using C++. The version now in beta runs on IBM's AIX operating system and has interfaces that support the Java and HTML languages.

IBM is considering licensing the technology to others, including some of its business partners such as i2 Technologies Inc. and Ariba Inc., both of whom have been granted access to technology in IBM's patent portfolio in the past.

Join the newsletter!

Error: Please check your email address.

More about Aribai2i2 TechnologiesIBM Australia

Show Comments