By Chuck VanDam, Supply Chain Engineering Manager, Agilent Technologies

In the fall of 2000 Agilent was faced with one of the steepest ramps in demand for test equipment ever encountered in the semiconductor industry. As markets heated up, unforeseen supply constraints pushed out standard component lead times by as many as 50 weeks, leading to expediting costs and gaps in supply that cut into both margins and revenue. But more damaging yet were the lurking liability “time bombs” created by ballooning material planning horizons during this period of rising demand. Purchase commitments made 12 months in advance to support up-and-to-the-right projections were suddenly converted into years of excess inventory when demand dropped off, followed by substantial balance sheet write-downs in the face of dwindling revenue.

While everyone at Agilent wants to avoid a repeat of this disaster, our experience has been that both the underlying problem of inadequate supply risk and flexibility management and development of an effective solution are instinctively relegated to the “procurement” function. However, as we have dug deeper into the matter, Agilent has concluded that preparing our supply chain for the volatility inherent in our demand environment must be approached as a set of interdependent business processes spanning sales, marketing, planning, finance, manufacturing and procurement in order to be successful.

The fundamental problem, as we see it, is that today’s tools and processes all center on collecting, processing and exchanging information that is primarily centered on preparing for a single demand scenario outcome. Agilent has long since concluded that much of the uncertainty that we are faced with in the management of our supply chain is here to stay – that is to say, that no amount of IT connectivity or forecasting wizardry is ever going to eliminate uncertainty about future demand. However, with the exception of standard inventory buffering algorithms built into or on top of our MRP systems, the types of data that we collect, process and exchange are simply not aimed at getting our arms around the range of potential demand scenarios that our businesses could encounter. We note, for instance, that once the “buffering” part of our planning process is complete, all the information that leaves Agilent to our suppliers goes out in the form of a single point forecast stretching over as much as a 24 month horizon, stating implicitly that there is really only one fixed demand scenario to prepare for.

But making the changes necessary to move from preparing our supply chain for a single outcome to explicitly preparing our supply chain for a range of potential outcomes involves making wide-ranging changes. For instance, we now recognize that, while demand volatility cannot be eliminated, valuable information about the potential range of demand is lost or “squeezed out” the very moment that point forecasts are “decided” and passed along to procurement. As a starting point, therefore, Agilent is moving away from projecting point forecasts, to projecting range forecasts that capture and quantify information about the limits of our uncertainty.

However, making this switch is not trivial. Marketing has traditionally been responsible for developing forecasts in Agilent. But since every veteran of the last several business cycles has long since concluded that accurately forecasting the future is futile, marketing tends to accept little responsibility for the quality of point forecasts that they generate, openly discounting their quality before they are even issued. However, with the introduction of range forecasts, marketing has a rich vehicle for expressing and quantifying both what they do and don’t know about future demand. Suddenly all of the old reasons for giving the forecasting process short shrift disappear, and new metrics for ensuring range forecast accuracy can and must be introduced.

Figure 1: A typical material requirements range forecast (in this case, for PCA assemblies) showing historical demand and projected ranges (10th, 50th and 90th percentiles of expected demand) over an18 month horizon. By definition, a “perfect” range forecast is expected to capture actual demand 80 percent of the time at any given time horizon.

Similarly, our traditional supply agreements do not adequately prepare our suppliers for a range of potential demand scenarios. For instance, in many cases nominal lead times are specified without agreeing upon meaningful volume limits within which lead times are guaranteed to stay fixed, implying, altogether unrealistically, that no supply constraints exist. It therefore also follows that the logic within our MRP systems makes the erroneous assumption of infinite supply at lead-time.

But here, once again, the changes required to actually account for, and then proactively modify the supply constraints that exist, require cross-functional involvement that goes well beyond the traditional boundaries of procurement. Working with our suppliers to ascertain what the constraints to our supply actually are is not just a procurement function, because in fact these constraints, while real, can be changed proactively, depending on the price we are willing to pay and the commitments we are willing to make as a buyer. At Agilent we now engage in a practice called “structured contracting” where we effectively define and secure supply “options” from our suppliers, effectively guaranteeing Agilent supply at fixed lead times and prices within certain volume boundaries. Making these options decisions optimally, however, involves range forecast inputs from marketing and planning, input from finance on the relative costs of shortages and inventory, and input from Sr. managers about our business’s tolerance for risk, since the risk of shortages and or excess inventory is largely a function of where availability, liability and lead time boundaries are set. Many functions that have traditionally, at best, thrown information over the wall have critical roles to play in the management of supply risk and flexibility, as illustrated in figure 2.

Figure 2. Cross-functional roles and responsibilities associated with supply risk and flexibility management at Agilent.

Perhaps most surprisingly, even to us, has been the fact that, of all the stakeholders that are impacted by the deployment of new supply risk and flexibility management practices, the easiest to move in this direction are, in many cases, our suppliers. Whereas demands on marketing, for instance, although more realistic, are nonetheless perhaps increased, supply risk and flexibility management techniques involving the use of structured contracts (known as StAR agreements in Agilent) to effectively place boundaries around the range of demand that must be prepared for, are almost universally welcomed by our suppliers. Typical comments include:

From a PCA manufacturer:
“ We are extremely pleased with the performance of the StAR agreements. We would not have been able to support the recent ramp without the upside requirements you presented in the StAR agreement.”

From a manufacturing of mechanical assemblies:
“ StAR is a good program and we would like to extend it to other boards. We have found that it is more difficult to do pricing, but it's worth it.“

From a power supply manufacturer:
“ This is the smoothest contract we have going. It's predictable, and we're willing to commit per the terms.”

Interestingly, therefore, most of the change involved in supply risk and flexibility management takes the shape of cross-functional process re-engineering. Indeed, it has been our experience that if supply risk and flexibility management is not seen as a “business” discipline, but is instead regarded as a “procurement” discipline, it will never actually get off the ground.



 

 

 

©2004 ChainLink Research, Inc.