In 2000 Hewlett-Packard faced a supply crisis. The flash memory used in its highly profitable printers also was being used by fast-growing cell-phone makers and HP was unable to obtain sufficient supply to meet its demand. As a result, the company failed to ship about 250,000 printers, which translated into a revenue loss in the tens of millions of dollars. Moreover, to ensure supply HP was forced to sign a three-year fixed-quantity, fixed-price contract with significant non-performance penalties for what historically has been a highly volatile component.

This experience served as the catalyst for HP to develop a framework for measuring and managing supply chain risk. Known today as Procurement Risk Management (PRM), the framework, which encompasses both process and technology, has become a key initiative at the company and is being applied to about $6bn of spend annually across numerous HP business units. Incremental savings to date total more than $100m.

The task of developing the risk management solution fell to a research and development team led by Venu Nagali, distinguished technologist at HP. "One of the first things we did when we started this program was to look out and see how other manufacturers were managing supply chain risk," says Nagali. "To our surprise we found that few manufacturers had any processes at all in place." Indeed, he adds, the team found numerous high-profile cases where failing to manage supply risk had resulted in significant losses of revenue and of shareholder value. These failures fell into three primary categories: price risk, demand risk and availability risk. "These three-demand, cost and availability-are the key uncertainties that have to be measured and managed," says Nagali.

All industries experience these uncertainties, but they can be especially volatile in the high-tech sector where HP is the biggest single purchaser of many components, including memory, hard disk drives and LCD panels. "It is good to be number one in any given market, but that also forces you to take on substantial risk, just by virtue of the large volume of components you have to buy," says Nagali. Additionally, high-tech components are among the most volatile in terms of cost and availability. Nagali offers an example of DRAM memory used by HP, which dropped in price by more than 90 percent in 2001 only to more than triple in 2003. "Except for the electricity market," he says, "I have not seen a market with more price volatility than high-tech."

Availability also is often uncertain. In addition to general disruptions caused by natural disasters or political acts, Nagali notes that in periods of high demand, suppliers place original equipment manufacturers like HP under allocation "whereby they supply only a fraction of an OEM's total demand."

The final variable, demand, is intensified by high-tech's notoriously short product life cycles. "If your product is successful, it lasts for six to nine months," says Nagali. "If it is a failure, it can last less than a month."

Wall Street Model
Since Nagali and many of his team members had prior experience working with risk management in financial markets, they studied the Wall Street model carefully. While many of the principles were applicable to the supply chain, the underlying stratagems used on Wall Street were not easily transferable. "Financial engineering practices enable the management of cost uncertainty through such instruments as call and put options," says Nagali. "But such instruments are not available for high-tech components." Financial risk management also fails to address demand and availability uncertainties, he says.

Conversely, current supply chain management practices emphasize the management of demand and availability uncertainties through inventory buffering strategies, but have little if any focus on managing component cost uncertainties. "The philosophy seems to be that all the risk can be pushed onto suppliers," says Nagali. "But what is not really well understood is that suppliers charge a premium to manage that risk. This is a hidden cost."

The bottom line, he says, is that in the supply chain, demand, cost and availability uncertainties all are equally important and are correlated. This means that they all need to be managed together.

"What we had to do was invent a completely new framework for managing risk," says Nagali. "It was not possible to simply take the math from Wall Street and apply it." This framework also required new technology, since existing supply chain management and enterprise resource planning software do not support risk management.

The first challenge the team tackled was how to measure and account for uncertainty in each of the three areas: demand, cost and availability. To do this, they developed new algorithms that are embedded in a software solution called HPHorizon. This solution begins with a typical point forecast, which it analyzes and corrects for biases. Then, using historical forecasts and current demand trends, the software builds statistically significant scenarios that calculate how high and how low demand could go, attaching to each scenario a probability number. Generally, the low number defines the 10th percentile, where the chance of demand falling below this level is only 10 percent. The high number defines the 90th percentile, which has only a 10 percent likelihood of being exceeded. Between these two numbers is an 80 percent range that is expected to encompass most demand.

The same type of tool is used to develop a high, low and middle range for the cost variability of component parts, typically looking six months out. "We have developed proprietary analytics that model the unique cost dynamics of high-tech components," says Nagali, "but these analytics also could be adopted to forecast the cost uncertainty of other manufactured commodities, such as plastics, chemicals or steel."

Determining availability uncertainty has been more difficult to automate. "What we are trying to gather from this scenario is the likelihood that we will be able to meet our demand without any contracts-just by going to the spot market," says Nagali. Unfortunately, he says, there are not enough data points to build a model at this time, so HP determines this variable through interviews with market specialists.

Measure, Then Manage
Once uncertainties have been calculated, with probabilities assigned, the question becomes how to use this information to make decisions differently, says Nagali. "The answer to this question is the more powerful aspect of PRM and it involves developing a portfolio of procurement contracts designed to share risks with suppliers," he says. "The risk sharing nature comes about by HP taking ownership of the risks that we can bear more cheaply and asking suppliers to take risks that they can more easily manage."

For the segment of demand where uncertainty is low, for example, HP enters into fixed-quantity, fixed-price contracts for a reasonably long period of time. "This is different to what manufacturers do now, which is simply to send a full point forecast to suppliers-just a forecast, not a commitment," he says. By just sending a forecast, "manufacturers are pushing the entire demand risk onto suppliers, which is unfair," says Nagali.

Using PRM, "what HP now does is to say to suppliers, 'here is a segment of demand that HP can absolutely commit to, guaranteed,'" he says. In return for that guarantee, suppliers are willing to give HP a price discount because the commitment allows them to manage their capacity more efficiently. "Committed volumes can be scheduled during non-peak times, and inventory carries no risk," says Nagali. Additionally, on high-volume deals, suppliers can modify fabrication lines to significantly reduce costs. "One printer supplier modified a conventional process based on HP's binding, forward commitment," says Nagali. "The resulting cost reductions yielded an additional 15 percent savings to HP, over and above volume discounts."

Less certain demand is satisfied through flexible quantity agreements. "Flexible agreements are the most common supplier arrangement in our industry, so creative modifications of these agreements are usually easy to pull together with suppliers," says Nagali. Suppliers often provide pricing discounts for committed upside volumes, he says, especially when the volumes have growth potential. Discounts typically increase as more volume is purchased. Making some of these commitments binding eliminates supply risk and provides further cost savings. "A significant percentage of HP's memory requirements are met through these binding but flexible agreements," says Nagali. "Contract horizons generally match HP's product lifecycle times and/or supplier capacity lead-times. The longer the horizon, the deeper the price discounts and the more binding the supply commitments."

For demand with high uncertainty neither HP nor the supplier make commitments. "Demand that is least likely to materialize can often be satisfied through the open or spot market," says Nagali. "As these sources dry up, secondary-sourcing options can be used such as brokers, auctions and product recycling programs." These approaches mean higher prices, but are often a better solution than carrying inventory, he says. And, the supply risk associated with these approaches often is less than expected. In one instance, HP's customer support teams realized significant inventory savings by recovering critical parts from unsold products. They also found consistent supply for low volume microprocessor demand through auctions, saving on inventories subject to severe price erosion, he says.

Optimizing the Portfolio
The next challenge for the PRM team was to help users determine an optimal contract portfolio for enabling specific business objectives. To meet this goal, the team developed additional new software called HPRisk, which provides contract valuation analytics. Since many contracts are in place at any given time, the HPRisk engine must look not only at current demand, price and availability of components, but also at the specific structure and terms of other contracts in effect, says Nagali. "Effectively, this software builds scenarios that show you how your world would look with one set of contracts versus another set of contracts. Then it helps you determine which set of contracts best meets your objectives," he says.

Objectives vary from one business unit to another, Nagali adds. "In some business units cost saving is the biggest objective, while it others it may be assurance of supply or cost predictability. "Using the PRM portfolio approach, we can enable any combination of objectives," he says.

Innovations in the HPRisk suite of PRM software have resulted in five patent applications thus far, but Nagali emphasizes that the solutions are very user friendly. "All of the math is hidden," he says. "The user just has to click on a few buttons."

For Nagali, "coming up with the math and embedding it in the software was the easy part of this project." A much bigger challenge, he says, was the people and process issues around getting PRM embedded in the HP organization.

Adding to this challenge, he says, is the fact that few employees, particularly in the supply chain, have any training in risk management, nor do they possess the skill sets or concepts needed to effectively manage this area.

Further, existing ERP and SCM systems do not support decision-making that is future looking. And metrics for assessing risk management do not exist. "There are metrics for almost every other thing, but not for supply chain risk," says Nagali. "Most metrics in the supply chain are about how long or how many, not about how much risk I am taking over the next six months."

Finally, the PRM process is extremely cross-functional. "For risk management to work, you need to have procurement, finance, sales and marketing, and the supply chain all working in vertical alignment," says Nagali. "Each has different information that is important to the process." At HP, he says, "we now have a cross-functional process where business-unit teams routinely look at risks and ask themselves, how are we going to manage this?"

The cross-functional process at HP is simple in structure but rigorous in execution. "Strategy and governance for a particular commodity typically includes approving procurement objectives, establishing metrics and reviewing performance of any existing portfolio of deals," says Nagali. Strategy and governance for product specific commodities is managed at the business level, while commodities common across products are managed more centrally.

Nagali's team serves as a consultant to introduce HP business units to PRM, to help them develop pilots and to train people on the software. More than 750 HP professionals have been trained in PRM concepts, says Nagali. Typically, it takes a business unit one year to two years "before they are ready for us to totally get out of the way and go it alone," he says. "At that point we give them the software and turn it over."

Range of Benefits
During the past five years, HP has implemented PRM for key strategic commodities as well as for procurement of certain indirect materials and services like advertising. "This wide range of applications illustrates the power and generality of the PRM approach," says Nagali. The process has enabled more than $100m in incremental savings, or savings over what would have been achieved through other methods without PRM.

Benefits can be categorized into four areas:
• Material costs savings: HP has obtained incremental material cost discounts up to 5 percent for standard components and an even higher discount for custom components, indirect materials and services. These discounts are a result of HP's quantity commitments, which lower supplier demand risks while enabling more efficient planning and production processes.

• Cost predictability: By including specific pricing terms in contracts, including price caps and floors where indicated, HP has been able to proactively manage cost uncertainty and protect margins.

• Assurance of supply (AoS): Managing component demand and availability uncertainties is a key objective for PRM at HP. PRM deals have improved AoS for several commodities even under conditions of an industry-wide shortage. For example, during an industry-wide shortage for memory that occurred in 2004, the PRM deals executed by a particular HP business unit ensured that they obtained 100 percent of their demand.

• Inventory cost reductions: The precise measurement of demand uncertainty using PRM software enables HP to optimize inventory levels internally and externally at supplier sites. Such optimization has cut inventory-driven costs by several percentage points for commodities implementing the PRM framework.

Suppliers also have benefited, says Nagali. "The quantity commitments that HP makes to suppliers, as opposed to just exchanging non-binding forecasts, has lowered supplier's demand risks," he says. "Suppliers of several strategic commodities have locked-up a significant portion of their capacity through PRM deals with HP."

This cascades through the supply chain, he says, with some suppliers making commitments to their suppliers based on HP's quantity commitments to them. The result is a significant drop in order volatility and a reduction of the "bullwhip effect."

Today PRM is considered a competitive advantage by HP senior management, says Nagali. "HP has an ambitious program to continue leading the industry and setting the industry standards in this important new business discipline."