The Defense Logistics Agency is the Department of Defense’s largest logistics combat support agency, providing worldwide logistics support in both peacetime and wartime to the military services as well as several civilian agencies and foreign countries. DLA manages nine supply chains and 5 million items, including clothing, subsistence, medical and spare parts. In FY2012, DLA had $53bn in sales and revenue, which would rank it 53rd on the Fortune 500 list.
Each fiscal year, DLA’s director provides guidance for determining where to focus resources and efforts in the upcoming year. For the past two years, these directives have been to reduce procurement requests by 25 percent—while maintaining or increasing material availability—in order to reduce costs through disposals and right-sized buys, yet improving support for the warfighter. To reflect declining defense budgets, DLA is also seeking to reduce its inventory by $10bn within the next five years.
DLA manages more than 1.4 million consumable Class IX items (repair parts), with $14bn in annual sales. The demand pattern for DLA items can be characterized as either frequently or infrequently demanded. Many DLA-managed items have highly variable demand, and managing them can be especially challenging.
DLA’s Enterprise Business System (EBS) uses demand planning software widely available commercially. This suite of tools has a number of direct impacts, affecting the magnitude of DLA’s inventory investment, number of procurement requests, and number of unfilled orders. It is not unusual for the variance of a DLA replenishment item’s monthly demand to be 10 to 100 times its average demand. Compare that to commercial inventory systems, where a variance twice as large as the mean is considered very large. Consequently, DLA’s commercial supply chain software that determines the timing and size of buys is faced with variability well beyond for what it was designed. The assumptions underlying the buy logic in that software do not fit the data, leading to buying the wrong items at the wrong time. This in turn leads to excessive inventory for some items, back orders for others, excessive procurement workload, and depleted working capital, says Robert Carroll, a DLA Logistics Operations J331 employee.
For items with frequent demand, EBS uses a suite of forecasting methods, ranging from simple (naïve, mean, exponential smoothing) to complex (adaptive smoothing, regression-based weighted average of methods). Unfortunately, these current forecasting models often generate large forecasting errors and multiple procurement requests, which increase the procurement workload and inventory investment value. From 2004 to 2010, key DLA inventory metrics (inventory investment, procurement workload, and wait time) trended in the wrong direction.
The process by which maintenance activities generate demands for repair items is inherently highly variable. Furthermore, these items have replenishment lead-times that range from months to years and represent low-volume business for manufacturers.
When lead-times can last that long, a common factor is the significant variability in customer demand, stemming from two major sources of uncertainty in the maintenance world: not knowing which parts will fail and need to be replaced and uncertainty because of changing programmatic factors, such as operating tempo, force structure, maintenance programs, and maintenance doctrine.
Neither of these factors is easily managed or controlled. Any system that seeks to manage the Class IX supply chain must cope with this high degree of uncertainty and expect a troubling result: large errors in demand forecasts.
Further compounding the challenge is that DLA is under extreme pressure to make the most efficient use of its inventory investment and provide a high level of support to its maintenance customers, and ultimately to the maintenance of weapon systems. Current level-setting policies for managing Class IX supply chains do not support these items well and do not give DLA’s decision makers the flexibility to implement alternative inventory strategies that offer tradeoffs among inventory investment, procurement requests, and wait time.
LMI has a long history of providing contracted support to the Defense Logistics Agency. In this case, LMI helped DLA achieve its stated goals to develop a more efficient system for managing items with infrequent or highly variable demand through an innovative system that accounts for need and risk, without sacrificing mission readiness.
Traditional methods for managing items with infrequent or highly variable demand often suffer from extreme forecast error, and, as a result, generate excess inventory and poor customer service, leading to some embarrassing events. “Which episode of '60 Minutes' would you like to be on,” quips Tovey Bachman, LMI senior consultant.
LMI bypasses the intermediate step of attempting to forecast demand for these inherently “unforecastable” items and employs risk-based hedging strategies to set levels directly from the demand history and item characteristics. To help DLA achieve a more efficient inventory, LMI deployed two of its independently developed tools—Peak Policy and Next Generation Inventory Model (NextGen), known collectively as PNG. These tools provide users with the ability to make three-way tradeoffs among inventory investment, procurement requests, and customer wait time, depending upon their organization’s specific supply chain performance objectives, Bachman says.
The result is a significant reduction in procurement requests and modest improvements in inventory investment, both in line with DLA’s stated organizational priorities.
In January 2013, LMI implemented PNG levels for approximately 500,000 wholesale demand items, which represent $3.6bn in sales for DLA. If expanded to 800,000 items and with emphasis placed on inventory reduction, LMI could drive down DLA inventory by $1bn over three years, while simultaneously improving customer service and decreasing procurement workload. For a research and development investment of approximately $6.5m, the return on investment is 27:1 for the initial DLA implementation.
By bringing PNG to DLA’s inventory challenges, LMI has shown that when organizations are demand planning, says Bachman, one size does not fit all. An organization should manage its inventory based on demand type and variability not by the options available in its demand planning software, and these PNG methods are applicable to companies with similar demand patterns/
That's good news for Carroll, who says, “Our object is not to employ demand planners but to get the best results.”
Keywords: forecast accuracy, demand volatility, demand sensing, demand variability, demand planning, supply chain solutions, supply chain management