Simulating the Supply Chain

The increased level of competitiveness in all industrial sectors, exacerbated in recent years by the globalisation of economies, is pushing many enterprises to strive for further optimisation of their supply chains. As a result, supply chain simulation technologies once taken up by only the most mature of supply chains are gaining increasing acceptance.

According to Gartner’s research vice-president C Dwight Klappich supply chain management (SCM) has long looked for tools that could enhance decision making around complex processes and activities. “As a result SCM is years ahead of many other functional departments in exploiting advanced mathematical and simulation technologies to predict future states based on available, though often limited, information,” Klappich says.

Supply chain simulation uses advanced mathematical engines and algorithms to forecast the future and to optimise processes such as planning production in a plant, how best to route trucks to reduce costs or even where the best location for a distribution centre might be.

Simulation tools attempt to “simulate” process behaviours by iteratively running the mathematical models that imitate real-world situations.

There are a number of places in SCM where simulation can be used, from modelling production processes to determining bottle necks or other factors that might impact production in a manufacturing facility. Network design applications that help determine where best to locate facilities based on cost elements such as transportation and demand are another type of simulation growing in popularity.

While supply chain simulation is not new, the tools are getting increasingly sophisticated. Klappich says there are newly emerging systems based on agent-based technologies that are “way out there in sophistication”. One example of this is the robotic warehouse management tool being designed by Kiva.

“Agent-based technologies are embedded in each robot and are constantly responding to changes in demand such that the robots can self organise with no human intervention,” Klappich says. “Agent-based technology is a specific type of simulation where the “agents” are endowed with behaviours that can mimic how something or someone might react in the real world. The iterative simulation then lets the agents run and a model of what might happen in the real world is the result.

“A good example might be a traffic jam where an agent represents an individual forklift driver in a warehouse and as the simulation runs you can see how traffic jams behave based on the aggregate of actions by all the agents. Agent-based simulation works when it is near-impossible to develop a single model for how a process might occur and the interrelationships between independent entities (agents) are easier to model. The simulation engine then uses this to mimic real-world
conditions.”

Getting simulation right

There are disadvantages, however. Supply chain simulation technology is a highly complex tool that requires experienced and knowledgeable users who understand how simulation works, how to model real-world systems and then how to interpret the output, understanding the mechanisms that led to a particular result.

Collecting the right data for successful simulation can also be a major challenge. CAS AG president and managing director Asia Pacific Markus Hoffmann says the availability of correct data is essential to using the technology. “In the Australian retail landscape, for example, that data is available but in other global markets such as Europe it is much harder to get that level of information,” Hoffmann says.

There also tend to be change management issues. “It’s more a mental thing,” he adds. “Yes, it’s sophisticated technology, you need to load data and it needs big databases and fast servers, but you can buy that.”

Klappich says another challenge is also inherently human. Companies often get results that are counter-intuitive and it takes a lot of trust to believe that a simulation is more accurate than someone’s “gut feel”.

“The disadvantage is not with the tool but with the use of the output. These are pretty sophisticated tools and you need users that understand how these systems work and can explain why the systems output the answers they do. An old operations research axiom is that 60% of the time, effort and cost goes towards modelling a problem properly and if the model is not right the output could be invalid,” he says.

Successful simulation

In the case of Fonterra, one of the world’s largest dairy producers, achieving the level of data collection needed to effectively carry out supply chain simulation is proving challenging enough. It decided that real-time visibility across a supply chain that spans more than 140 countries is the first step to simulation.

To achieve this, Fonterra has built a global supply chain visibility solution using business activity monitoring technology based on Software AG’s webMethods suite. This is, however, only the first part of the process.

In future Fonterra intends to also perform predictive trend analysis/simulation across its supply chain. Instead of alerting on single instances, the system will be able to continually analyse data and alert if trends outside tolerance levels or norms are identified.

The cost of getting data and the sheer cost of simulation packages that can run to up to $1m means supply chain simulation and its subsets such as trade promotion simulation software tend to be the province of the big end of town, says Gartner research director SCM, Tim Payne.

“The exception to this rule is high-tech companies,” Payne says. “These companies are based on technology and with products going obsolete in six, 12 or 18 months, many of them almost can’t afford not to.”

While the cost of implementing supply chain simulation technology may seem stratospheric, the ROI many companies are deriving is equally impressive. Payne says there are large companies with revenue in the billions that have saved tens, if not hundreds, of millions through the use of simulations.

German biscuit manufacturer the Bahlsen Group saw a similarly impressive ROI when it implemented CAS AG’s trade promotion simulation technology. According to a Gartner case study (Ref: G00164772), forecast accuracy went from 50–70% forecast accuracy to above 90%. Because the simulation could accurately predict the volume that a promotion would generate, it was better able to drive increased revenue and profitability.

Other benefits achieved through the implementation included an estimated benefit of €1.1m annually from being able to avoid product out-of stocks that result from inaccurate forecasts, and cost avoidance of €30,000 annually in packaging destruction costs from seasonal products due to the inability to forecast demand.

“In the hands of a strong user these systems can often save companies large sums of money and I have seen numerous cases where the system was more than paid for after the first simulation was run,” Klappich says.