Software Peeks Into Purchasing

As e-catalogs proliferate, ePlus Inc. and Zycus Inc. are each readying software to make it easier to quantify and classify products bought and sold in business-to-business transactions.

ePlus is upgrading its Content+ online catalog engine next quarter with the capability to massage classifications beyond the more than 44,000 commodity types currently available in its hierarchical schema.

Because companies view products in different ways, new features in Content+ will let users organize various descriptions in a unified manner with standardizing templates developed by ePlus. The Herndon, Va., company provides tools that enable approved users to accept new products into the system, as well as change product listings and routings. The upgraded platform will also have a supplier portal that allows suppliers to add, delete and make price changes to products in a users Content+ e-catalog. This can be done in batch mode, individually or by role, as an administrative function, officials said. A Schema Management module will enable the system to validate an item when it is changed or added to the system.

Separately, automated catalog content solutions provider Zycus this week will announce AutoClass 1.5, classification software for e-catalogs that standardizes e-procurement content for exchanges and e-marketplaces.

AutoClass, which is part of the Santa Clara, Calif., companys E-catalogs suite, classifies products into one or more of 13,000 categories that conform to UNSPSC (Universal Standard Products and Services Classification) code, according to officials.

The classification capabilities of AutoClass should go a ways in simplifying catalog integration for buyers and exchanges, as it provides a common platform for mapping disparate supplier and buyer schemata to the UNSPSC schema, according to officials. AutoClass also supports additional schemata, including eClass, and can support customer-specific standards.

While UNSPSC is an emerging standard categorization structure for products and services, most suppliers and buyers participating in e-commerce initiatives still have their own categorization structure.

AutoClass automates UNSPSC-based classification through the Bayesian Inference Engine, proprietary artificial intelligence software that extracts and links product concepts and categories. The Bayesian Inference Engine can automatically classify 100,000 documents per hour, said Zycus officials.

Marc Thomas, general manager of Global Sourcing for General Electric Co.s GE Transportation Systems unit, in Fairfield, Conn., is using Zycus E-catalogs to figure out with a high degree of accuracy what GE is purchasing across its many units. Last year, the company spent roughly $50 billion on products and services, yet it didnt have a clear picture of what it was buying or whom it was buying from, said Thomas. To determine what it was buying, GE decided to use the UNSPSC schema.

"With the language of UNSPSC, we now had to find a tool that would take that code and standardize the application of what we bought," said Thomas.

GE is using Oracle Corp.s 11i iProcurement applications for requisitions. Since the system does not lock down UNSPSC codes, GE can determine the aggregated amount of spending that goes through the system but not what the money was spent on.

"We couldnt go back in the marketplace and negotiate contracts," said Thomas. "Now with the auto-classification tool in place, it gives you accurate descriptions [of our spending]."

At the same time, the Zycus tool increases the rate of accuracy when classifying spending, as it does not allow buyers to type in their own code for a product.

"[With Zycus], you type in what you want, and in 2 seconds, [the auto-classification system] comes back with a No. 1 choice. If youre an experienced buyer, you might say, This is not exactly what I want—the tool is only 85 percent accurate," Thomas said. "You can go back and hit a button, and it will pull up more options, so you cannot type in your own code."

GE is working with Zycus to classify the remaining products that do not have UNSPSC codes.

In starting the project, Thomas said GEs biggest technology hurdle was finding an application that could take on the classification job.

"[Procurement managers] were ordering from a catalog vendor that we had set up. Those guys would take anywhere from a few days to 60 days to turn around and classify [a purchase] based on our description from a catalog," Thomas said. "The supplier would either give you clean information or not clean information, and it was manual [to do the classification]. So what would take up to three weeks to do now takes a couple of seconds to do."

With the current Zycus system, Thomas has seen pretty good results, moving from 42 percent classification accuracy for catalog purchases and 40 percent for noncatalog purchases to 80 to 85 percent accuracy in both categories.