I may be somewhat unusual, but I've never bought anything at an online auction. I've seen eBay, and one of my friends sold some of his collection of valuable magazines (okay, comic books) on eBay, but I've never gone the whole route and come home with the goods. I've thought about it a couple of times. I recently built a computer from parts, and one of the places I looked for motherboards had an auction, but they didn't have the board I wanted and there was no way to post my own offer to buy.

Auctions consist of at least two different processes, if not three. There's the straightforward auction, in which a seller posts an offer of something that multiple buyers can bid on. Then there's the reverse auction, in which a buyer posts an offer to buy something, and multiple sellers bid progressively lower prices (at least in theory). And there's the true exchange, in which multiple buyers and sellers gather and a product changes hands whenever two or more participants can agree on a price for particular goods.

Of course, there are multiple variations on auctions. Things like English auctions, Dutch auctions, silent bids, and so forth. These all have more to do with the behavior of the auction than with the concept of the auction itself. Rules such as minimum bid intervals, who can see information, who can participate are all part of the package, but the concept stays the same: buy and sell.

A number of vendors produce software to implement auctions and exchanges, including companies like Trading Dynamics, Tradex and Tradium. You might have heard about Ariba buying both Trading Dynamics and Tradex. This industry is hot.

But it's also off target, at least slightly. I work in Net Markets, and meet with clients daily. I wish I had stock options for every time I've gone into a conversation with a potential new market that thinks they want auctions. They don't. Not really - or at least not only.

This is true for several reasons. Now, most of my conversations have been with business-to-business (B2B) clients rather than consumer-to-business (C2B) clients. On the consumer side, the auction model has more validity, because in general you're selling products with a relatively small price tag in relatively high quantity. But B2B is different. A consumer site like eBay may have millions of participants, but many business-to-business sites have audiences in the hundreds because they focus on either high unit price or high-volume orders. Only so many companies want to trade in airplane parts, for example.

That's not to say this isn't a good market to be in - a book may cost $5.99 on Amazon.com, so their model is clearly volume: making millions by moving a large number of small-ticket products. An airplane engine may be worth millions, so the number of deals will be smaller, but the overall dollar volume may in fact be much larger than most C2B sites.

And that's one of the things that drives B2B to need more than auctions. Remember, an auction is essentially an open market where the only relevant attribute of the market is price. Exchanges are slightly more complex, but the main element of an exchange is still price matching between the buyer and seller.

That's where the model breaks down in B2B. Most B2B sites aren't set up as a buyer or a seller, but as a broker. They take a commission on the sale - most times from the seller, sometimes from the buyer, occasionally from both. But because of the large price structure, and the nature of their products, an auction isn't always appropriate.

Most high-dollar products have multiple properties that affect the trade in different ways. For example, a product may come in several grades. An offer to buy might be for the higher grade, but in reality the buyer might be able to use the next highest grade should other options such as price make it attractive. So the participants in this kind of market would want to trade not only on price but also on grade. Then shipping, or location, or packaging or a hundred other factors may come into play. Technically, they want multiple attribute trades. And what they really want is negotiations.

An auction almost always has an automated clearing. The highest bidder wins when the auction ends. A negotiation doesn't necessarily end that way, because price is only one criterion for assessing the best deal. You might have to figure in freight charges, which may be lower if the seller is in the same state or higher if the seller is out of the country. You might be comparing two similar grades, or two grades that can do the same job but have different efficiencies that need to be accounted for. The rules are generally too complex to code to an automatic conclusion.

Other factors come into play. Auctions and exchanges are designed to be open to as wide an audience as possible because it drives liquidity. Business deals don't always work that way. A company that sells steel to General Motors for $80 a ton doesn't necessarily want GM to be able to bid on their steel at $60 a ton in an online market, or even to know that they are providing steel at that price. Anonymity is usually a requirement, but restriction of trading partners is also common. These tend to constrain a market somewhat, but since many of these markets are for multiple-attribute products, a commodity approach may not be appropriate anyway.

Unfortunately, at the moment there's a big, negotiation-engine-shaped hole in the market. There are no shipping products that handle negotiations; most really can't even deal with multiple attributes. The race to get there first is heating up; Ariba isn't the only company buying market product makers. Better get in the bidding fast for that first negotiation company, because it'll quickly be going, going, gone!