CombineNet: Moving Beyond Reverse Auctions
Using combinatorial optimization, DuPont finds further ways to save on transportation.
For forward-looking companies, "strategic transportation sourcing" is a big trend these days, says Pierre Mitchell, vice president of research for AMR Research, Boston. The movement has two components.
First, these companies have started to view goods and transportation as part of one integrated spend. How goods arrive is no longer an afterthought: procurement professionals consider the cost of shipping as part of the overall price.
Second, these shippers are cooperating better with transportation suppliers. Instead of dividing complex networks into arbitrary bundles of lanes to simplify bidding, they ask carriers to bid on the lanes where they can offer the best pricing and service.
"Michelangelo didn't build the David by cutting out lots of little blocks and subbing them out, then gluing them all back together," Mitchell says. "It's not a reductionist operation. You have to use a network-based solution."
To reach that kind of solution, you need a tool that's more capable than the standard online reverse auction, notes Mitchell. You need a tool that's based on "combinatorial optimization."
Chemical industry giant DuPont recently adopted just such an application for its large, complex transportation sourcing decisions. Since February, it has been using Pittsburgh-based CombineNet's Decision Guidance System for several major procurements.
For a while now, DuPont has placed computerized bidding at the heart of its overall procurement strategy, says Mario Hegewald, leader of DuPont's Global Logistics Center of Excellence. The technology, he explains, helps the company uncover the real market price for the desired good or service.
Using reverse auctions, "we've had tremendous success—not just in logistics, but in all our negotiations with suppliers—in driving costs out of the system," Hegewald says.
Logistics Center of Excellence
The Global Logistics Center of Excellence provides fourth party logistics (4PL) services to DuPont's 22 strategic business units (SBUs). Hegewald's teams in the United States, Europe, Asia, and South America purchase transportation for finished goods moving out of DuPont's facilities around the world. They also arrange transportation for many of the raw materials heading into DuPont's production facilities.
DuPont has standardized on a single bidding tool to handle most of its procurements. But the Global Logistics Center of Excellence has gone its own way in working with CombineNet. The system offers greater savings than other tools DuPont has used in the past to conduct transportation procurements and award lanes, Hegewald says.
Today, customers such as DuPont call on CombineNet to save money on products and services. But the company first developed its combinatorial optimization techniques for a wholly different purpose—espionage.
The problem was how to give various branches of the U.S. military, intelligence agencies, and other constituent groups the access they needed to spy satellites, which collected information only during limited hours.
"Balancing all those competing agendas and different privileges, and achieving optimum utility of the satellite, is a very difficult challenge," says Tom Finn, CombineNet's executive vice president.
Constraints with a Cost
A large procurement also balances competing agendas. Companies try to cut costs by aggregating purchases and looking for the lowest bids. But they must evaluate service quality as well as cost, and they must consider a multitude of business constraints based on everything from government regulations to company policies to managers' preferences. Each constraint carries a cost, Finn says.
CombineNet's technology allows users to run vendors' proposals through hundreds of what-if scenarios, to help reveal the costs of sticking with or dropping different business rules. The result: "Plant managers stop arguing that they're only going to do business with one particular supplier," Finn says.
Armed with a CombineNet analysis, a sourcing professional might, for example, show that adding a second supplier could save the company $1.5 million.
"When all the hard dollar tradeoffs are revealed, it doesn't just expedite the process; it changes the process," Finn says.
Because CombineNet's system quickly analyzes many thousands of factors, it can accept highly customized bids from vendors. "We went to CombineNet because it allows these bidding events to occur in a much more free-form fashion" than many procurement tools, Hegewald says.
Submitting Expressive Bids
Less advanced tools allow buyers to carve their thousands of transportation lanes into lots of 100, 1,000, or whatever they prefer, Hegewald explains. A carrier can bid on any of those lots, but it must bid on all the lanes in a lot, even if it can't serve all of them equally well.
CombineNet allows vendors to submit "expressive bids." They tailor these bids to their strengths and needs by including the unique conditions that make sense to them. For example, Hegewald says, an ocean carrier might propose: "If you give me this lane from Miami to Argentina, then I'll give you the China-to-U.S. lane for x discount."
CombineNet's system is not the only one that accepts and analyzes expressive bids, says Mitchell. But CombineNet has found great success in the market, and one of its customers has conducted one of the largest, most complex transportation procurements ever accomplished, he says.
As of August, DuPont had used CombineNet to award 12,000 ocean lanes and to conduct a procurement for LTL transportation. It was also in the midst of a procurement for all its air freight, and for truckload transportation in the United States.
As an application service provider (ASP), CombineNet usually hosts the Decision Guidance System, providing access to buyers and their vendors over the web. Customers have the option of licensing the software and running it behind their own firewalls, Finn says.
Before starting a procurement with CombineNet, "you've got to have the data," Hegewald says. Working with the SBUs, team members at the Global Logistics Center of Excellence collect information on transportation lanes, traffic volumes, and required equipment, among other factors.
They also gather what DuPont terms Critical to Quality (CTQ) criteria—business rules the system will use to evaluate bids. For example, he says, "Is a service level more important than price? Are there particular handling requirements that you need for a lane because it's particularly hazardous material?"
Uploading Left and Right
The team loads that information into CombineNet's application by entering data at the keyboard "and uploading Excel spreadsheets right and left," Hegewald says. Compared with other tools he has used, CombineNet's software makes it particularly easy to manage this data, he says.
Before launching the reverse auction, DuPont also pre-qualifies vendors and brings in the eligible ones to show them how to use CombineNet's technology. Then the bidding begins.
"If you're doing a true reverse auction, people can actually see what those bids are and watch prices start to tumble," just as bidders on eBay watch one another drive prices up, Hegewald says.
CombineNet allows vendors to see the conditions that others propose in their expressive bids, but DuPont chooses to hide that information; vendors see only their competitors' complete package prices.
When all bids are in, DuPont uses CombineNet's system to run numerous scenarios that test what would happen if the company gave more or less weight to various business constraints. Once Hegewald's team has tested every set of combinations it wants, it picks the solution it likes best and awards the lanes.
Since DuPont embraced electronic bidding for transportation, the procurement process "has sped up exponentially, from 12 to 18 months to six months, and now down to three months," Hegewald says.
CombineNet's system doesn't necessarily work faster than other computerized bidding tools, but "it's the right tool for complex events," he says.
Last, Best Hope for Savings
An application such as CombineNet's is the last, best hope for savings at companies that have already trimmed a great deal of fat from their transportation networks by other means, says Mitchell.
"The more advanced companies are using these combinatorial optimization tools just because there's no other way to manage all the possible permutations," he says. "You just can't do it with traditional RFQ tools. They will destroy value."