An advert caught my eye the other day regarding supply chain management solution – it read “Achieve a More Responsive Supply Chain”. The vendor was promoting the concept that manufacturing agility and responsiveness is something to be pursued; and implied that companies that can respond quickly are far more likely to balance supply and demand within their supply chain. Intuitively this does makes sense - increased manufacturing responsiveness can certainly lead to a number of benefits to the end customer, such as improved service levels, fewer stock outages and improved profitability. But is agility and responsiveness necessarily a good thing? In a complex system where multi company supply chains rely on many interlinked factors, speed and agility in one part of the supply chain may in fact be a destabilizing factor in other parts, and eventually lead to the total opposite of the intended results – stock outages, excess inventory and dissatisfied customers.
The well known author Peter Senge in his book “The Fifth Discipline” describes a simulation game developed at MIT’s Sloan School of Management in the 1960’s. This simulation called “The Beer Game” simulates the distribution of beer, from a brewery, through the wholesaler, to the retailer and finally to the consumer. This network is characteristic of many production/distribution supply chains, and the Beer Game illustrates several characteristics of a typical distribution network.
The Beer Game simulation starts with a steady state – where the brewery output is matched to customer demand. In this stable situation the orders placed by the retailer on the wholesaler, and by the wholesaler on the brewery are in balance – production meets demand exactly. Then (in the game) consumer demand for a particular brand is ramped up over a few months (as a result of a promotion) and is then ramped down as consumers switch back to their preferred brands. What is the impact of this on the overall supply chain?
In the simulation the retailer notices an increase in demand, and as his stock runs out he increases the size of his orders placed on the wholesaler. It takes time for the wholesaler to notice that increased orders from many retailers are resulting in stock outages in the distribution warehouses; and in turn to compensate for possible stock outages the wholesaler increases the size of orders on the brewery. However, owing to the lags in the system; as customers continue to buy more of that particular brand further stock outages occur at the retailer; and in compensation the retailer again increases the size of his orders to the wholesaler to ensure that he has sufficient stock. These increased orders reinforce the wholesalers’ perception that he needs to place an even increased order on the brewery. The brewery, noticing the trend of dramatically increasing orders makes plans to switch production to the brand that is clearly selling so well. Again, lags in the system mean that it is some time before production is switched and a massive production run of the particular brand is initiated to catch up the backlog. Of course, in the simulation; by the time the production is increased; the demand has dropped off as the effect of the promotion end and consumers switch to another brand. The upstream brewery continues to pump more beer into the network to fulfill backlogged orders; until all of a sudden everyone (except the end customer) is sitting with surplus stock.
The beer game simulation demonstrates that supply chains are systems with many interlinked components; and that the individual actions and the decisions taken by the retailer, wholesaler and the brewery, while perfectly rational when viewed in isolation can in fact cause a destabilizing effect on the upstream (and downstream) supply chain.
Engineers are used to modeling natural systems and accommodating time lags. For example the speed control in a car can increase the throttle gradually until the actual speed reaches the desired speed (set point). Such control systems are finely tuned to account for the time lags in measurement and the dynamic response of the system itself. The throttle is therefore gradually increased to present overshoot. In chemical plants, process controllers utilize PID (proportional, integral and derivative) algorithms to account for lags in the system. For example the pressure control on a distillation column can adjust column pressure by opening and closing a valve in the gas vent. An engineer will confirm that incorrectly changing the tuning parameters can rapidly result in instabilities and pressure swings in a distillation column can be extremely difficult to control once the system is unstable.
With typical supply chains we have a system that is even more complex and unpredictable than the average PID control loop on a chemical plant. In a distribution supply chain measurements are not easy – the units of measure can be batches, shifts, days or even months with significant time delays. In a distribution supply chain the human factor also introduces new uncertainty – the manager of the retailer, wholesaler and brewery each thought that they had made the right decision to increase orders or production – but they all over compensated. The result is that the distribution supply chain went unstable, and stock outages rapidly turned into a surplus. Had the individuals understood the impact of their actions on the rest of the supply chain they may have moderated their response, and the overall system would have been more stable.
Supply chain solutions that increase the responsiveness and dynamics of one part of a supply chain without understanding the impact of this response on the rest of the supply chain can actually make things worse. If we understand a more “responsive” supply chain to be the orchestrated response of all elements of the supply chain to any change in the system then we are on the right path. If however we interpret “responsiveness” in isolation (for example increasing production without taking account of distribution), or increasing stock radically in response to increased customer orders; we may in fact be destabilizing a system that was possibly working close to optimally before.
The solution is to consider every supply chain initiative within the context of the whole system. With the “balanced orchestrated” approach it may be more prudent to slow down change and accept short term imperfections – while giving the whole supply chain a reasonable chance to adapt incrementally to the change in demand or supply. Supply chain modeling needs to be done at many levels; at the lowest level it involves production scheduling. At the highest level demand planning and forecasting is important – and it is at this level that the most collaboration between participants in the system is required. An orchestrated approach demands that all levels in the supply chain are considered, both within your company and in the other companies that participate in the same distribution network.
While the supply chain solution advert may have been promising a faster, more rapid response; a better approach is surely a more measured response through improved insight into the dynamics of a specific supply chain? The tools and techniques and experience are available to achieve this and every company in a manufacturing / distribution network should be considering how they can achieve this enhanced understanding and insight into the dynamics of their own supply network.