For someone who has been working in and around the supply chain for more than 25 years, Lawrence "David" Davis sometimes sounds, well, not like a supply chain kind of guy. He talks about "genetic algorithms" and "evolutionary computation" and "deterministic simulation," which at first blush might seem to have little to do with running a company's supply chain.
But Davis, who is a senior fellow with Charlotte, N.C.-based NuTech Solutions, is one of a group of supply chain thinkers on the forefront of applying advanced science to solve thorny supply chain challenges. And one of the thorniest at present, in Davis' view, is that 20th century supply chain solutions are not up to the challenges that companies face in managing their supply networks in the 21st century.
Those challenges include supply networks that are becoming increasingly extended and complex; integration between companies and their trading partners that is becoming deeper at the systems and process levels; and emerging technologies like radio frequency identification that are producing ever-growing mountains of supply chain data. These and other factors threaten to overwhelm the systems that companies rely on to monitor and manage their flows of goods. Moreover, 20th century systems may be inhibiting companies from moving toward a 21st century supply chain.
Debugging the Supply Chain
Davis sees several problems with current supply chain technologies. First, he believes that contemporary solutions do not allow companies to optimize at the appropriate level of aggregation. By that, he means that companies should be able to use solutions to optimize across their sourcing and procurement, production and distribution processes all at the same time. Instead, supply chain solutions tend to break those functions out into separate modules, each of which runs separately from the others. "The answer you get when you first do production and then do distribution or any of these functions independently is not as good as the answer you get when you do them together," Davis says.
Second, software solutions typically optimize based on deterministic assumptions about how long it will take for any given process to be completed. As a result, the solutions produce "perfect" schedules that do not allow for breakdowns of machinery, traffic jams, defective parts and other real-world circumstances. "If you have a schedule that is built on the assumption that all your assumptions will be correct, the failure of any one assumption can make the whole thing fall down like a house of cards," Davis says.
To produce a plan that can be robust in the face of the many possible failures of assumptions, Davis asserts that supply chain solutions must employ stochastic simulations that simulate the effects of plans and schedules again and again and again, running through a series of "what if" scenarios to determine how well a particular schedule will perform against the range of possible futures. The final plan might sacrifice a small degree of optimization that a company would never be able to capture anyway, but the plan would be able to withstand inevitable changes in assumptions.
Third, Davis believes that these kinds of stochastic simulations must have embedded agents that follow the company's business rules. "At every point in a stochastic simulation when things go off the rails, we need to refer to the business rules of the enterprise in order to ascertain what the effects will be of a deviation from our expectations," he says.
For example, if a dispatcher is expecting to consolidate three loads at a dock in Chicago and one of the three loads winds up being late, that dispatcher likely has a set of procedures that he or she follows to decide whether to send the consolidated load on minus the late one or hold the shipment and send all three later, with all the subsequent downstream effects on the schedule. A simulation of that series of events should take account of the company guidelines that the dispatcher would logically follow.
However, Davis points out that by embedding intelligent agents in a simulation that is reacting to stochastic events, when those agents provide feedback that affects the outcome at different points along the way, the simulation spins out of predictability as minor adjustments to schedules get magnified over time. Therefore, to come up with an optimal plan, a 21st century supply chain solution should be able to run through the simulation over and over to determine the impact of each change in circumstances. The end result of the simulation would not be a single fixed answer but would be a distribution, a range within a standard deviation that would allow a planner to select an outcome based on the relative likelihood of that outcome actually coming to pass and the planner's — or the company's — willingness to accept a specific degree of risk. "If someone asks how many units we'll sell next month, and someone else gives an answer of 50,000 units, that answer is always wrong if the question really relates to a process where there is lots of stochasticity," Davis explains. "The real answer to the question is to show a distribution. It may be centered on 50,000 units, but you need to see the whole distribution."
The Future is Now
Davis has been putting these theories into practice both at NuTech Solutions and at a NuTech spin-off called VGO Associates (of which Davis is president), serving clients like Air Liquide, a producer of industrial gases. For Air Liquide, NuTech used so-called genetic algorithms and ant algorithms to determine the best production schedules and distribution points across a supply network that involved 40 plants and 8,000 client sites. The optimizer created for this project took account of such factors as power prices and customer demand projections, daily power costs and efficiency for every plant, production costs based on forecasted demand, and potential maintenance and power issues at each plant. While Air Liquide does not publicize the full impact of the new optimizer, the company has enthusiastically embraced the solution and touted its contribution to helping the company make more profitable daily operational decisions, yielding significant savings while improving customer responsiveness.
When will these kinds of capabilities hit the mainstream? No time soon, at least in the solutions offered by the major supply chain software companies, according to Davis. Niche companies like NuTech and VGO Associates already offer many of these types of capabilities, of course, and leading-edge enterprises are using solutions from these firms to tackle some of their toughest supply chain optimization challenges. But Davis believes that a truly 21st century supply chain solution will evolve from the conglomeration of various point solutions into a new type of supply chain planning system. "I'd be willing to bet you decent odds that in ten years' time there will be a supply chain solution that embodies all the things I've talked about," he concludes. In the meantime, companies looking to embrace complexity and take advantage of the new science of supply chain will have to settle for customized applications if they want to plan globally and optimize genetically.