by Jamie Beckett
In early 2000 when the world was intoxicated with online
auctions, Kemal Guler wasn't among those rummaging through
his garage for treasures to sell on eBay. Guler, an economist
at HP Labs, was instead thinking about how the Web would
change auctions forever.
Auctions have been around for about five millennia, selling
to the highest bidder everything from horses to silver to
securities to works of art. But electronic auctions open
up a range of new possibilities – including the opportunity
to easily experiment with different auction formats. Suddenly,
almost anyone could run an auction, or participate in one,
and all sorts of goods – from Victor gramophones to
an entire plastics factory in Mexico – were attracting
bidders from all corners of the world.
"An auction house like Christie's or Sotheby's had
a fixed set of rules and applied it in auction after auction
after auction," Guler says. "With the Internet,
we were in entirely new territory that would create new forms
of auctions that had never before existed."
Guler set about exploring that new territory as part of
a larger effort at HP Labs in business-process innovation,
a means of finding smarter ways to manage commonplace business
tasks such as product portfolio optimization, supply chain
planning and, in this case, business procurement, which increasingly
involves auctions.
Over the past four years HP has used online bidding to purchase
some $21 billion worth of materials – items like power
cords, memory or packing materials – at an estimated
savings of $1 billion so far.
Overall, U.S. businesses are expected to use online auctions
for more than $660 billion worth of procurement operations
in 2006, according to a recent IDC report.
"An auction allows you to quickly identify the lowest
cost you can achieve in certain conditions," says Karim
Mohara, HP's manager of e-sourcing. "When you negotiate
one-on-one, there is always room for uncertainty. It's like
buying a car: you always wonder if you got the best price."
For HP, the goal was not just to reduce costs or speed up
the procurement process. It was to use auctions intelligently
and to do so without putting a strain on its supplier relationships.
"Today people configure auctions randomly or intuitively," says
Mohara. They make ad hoc decisions about such auction attributes
as whether bids will be open or sealed, the order in which
items are auctioned, lot sizes and the amount of information
buyers receive.
But ‘ad hoc’ auctions aren't necessarily the
most successful. Changes in auction format and bidding rules
can substantially affect auction outcomes.
Electronic auctions, says Guler, make it easier to change
these rules, to set up more complex auctions and to collect
data on the results.
"In a traditional auction, you were restricted
in that everyone had to be present," he says. "More
importantly, the way bidders expressed what they were
willing to pay had to be simple because all that information
had to be processed very quickly by one fast-talking
guy."
In an electronic auction it is possible, for instance,
to allow bidders to offer different discounts for different
quantities.
"If you take bids for 10 different quantities from
10 bidders, determining how much you purchase and from
whom is not something you can figure on the back of an
envelope," says Guler. "Traditional 'price-only'
bidding languages, in their beautiful simplicity, are
simply inadequate for the kind of communication needed
for complex business cases."
HP has put the same tool to use when it's been on the
opposite side of the auction block – bidding to win
business from a prospective customer. In one instance,
a large telecommunications firm held a reverse auction
(in which sellers lower their prices until the buyer accepts
one) to purchase workstations, desktops and notebooks.
The result: HP not only beat rivals, but the company won
the $15 million contract at a higher-than-anticipated profit
margin.
Now, researchers are exploring applying this knowledge
to HP's overall pricing and sales policies.
HP is also putting its expertise to work for selected
partners. Researchers' auction analytics work has shown
a major car manufacturer how it can reap more revenue from
wholesale car auctions to dealers by adjusting the order
in which cars are put up for bid.
By adjusting this one factor, the car company can boost
the sale price per car by one percent, the probability
of a sale by three percent, and auction revenue overall
by four percent – and pocket $160 million a year
more in profit.
So far, HP has 20 patents pending on various aspects of
auction analytics.
Researchers have created an early version of software that
can help procurement professionals design better auctions.
They're now working with the eSourcing team to collect and
analyze more data so the tool can be customized according
to specific auction details.
"Our dream is to make auction decision support a more
automated process," says Guler.
Long term, Guler says, researchers want to apply their skills
in game theory, economics, econometrics, statistics and decision
analysis to a whole host of business processes.
"There are volumes of scholarly literature on these
topics and, at the same time, there are emerging business
needs," Guler says. "We want to fill the gap, to
bring science to business decisions."
Jamie Beckett is managing editor of the HP Labs Web site.
Before joining HP, she was a reporter and editor at the
San Francisco Chronicle. She is also a published fiction
writer.
|
Printable version
