Managing spare parts inventories for military equipment is challenging. Spare parts demand is driven by maintenance customers, but predicting when and how often those needs will occur is impossible. Most complex systems are maintained in a turbulent environment, driven by unpredictable human decisions that drive budget crises, wars, reset, maintenance programs (fix this, not that), operating tempo, deployment, and so on. On the supplier side, manufacturers and vendors see buys for new stock or repaired items as unpredictable and erratic. It isn’t steady business, so suppliers work these items into production at their convenience. This results in lead times of months to years.
Assumptions and Supply Chain Planning
When people modernize inventory planning systems, they assume that replacing old systems will drive improved business results. Sales pitches conflate getting a shiny new software system with improving quantitative results for your business time and time again. While the pitch may be intriguing, it doesn’t necessarily mean it’s true. Systems are built on assumptions, and when assumptions are wrong, so are the results.
Planning software assumes that demand fits statistical models, or that customers collaborate and provide notice of demands. But human decision makers don’t behave in accordance with predictive analytical models, and they don’t notify maintenance personnel of decisions they don’t yet know they will make. Therefore, planning assumptions about maintenance customers’ demand data are wrong. When it comes to suppliers and manufacturers, planning systems assume lead times are short. It’s also assumed that suppliers are agile, and therefore demand won’t change much during the wait for replenishment. But this isn’t true either—human decisions can completely change parts demands over lead times of months to years.
Use a Robust Strategy
What do you do if you can’t predict something? You hedge your bets; rather, develop a strategy that works well under a variety of possible future scenarios. That’s exactly what LMI does with its Peak and Next Gen (PNG) inventory management solution. Customers send us data for a wide span of time—typically 5 to 8 years—and our team uses data containing ripple effects of all those unpredictable human decisions. That data supports a sophisticated mathematical hedging strategy that stress-tests candidate stock levels against many possible demand transaction patterns and determines which perform best. Decision makers see tradeoff curves that they use to align the strategy with their objectives for customer support, inventory, and buyer and repair workload. When they select a point on a curve, it determines the PNG settings, which in turn determine stock levels for every item. Customers then load those levels into their planning system.
Putting It to the Test
LMI understands the unpredictable spare parts environment because we’ve spent years working with customers who have sporadic or highly variable demand items that play havoc with forecast-based solutions. This led to the key insight of recognizing that forecasting parts demand is futile. For example, a major contractor implemented an ERP system and sophisticated supply chain software at a DoD agency. After 6 years, all key business metrics had worsened; inventory was up, customer service was down, and procurement backlogs built up to more than 7 months. DoD decided to move its most difficult items to LMI’s PNG solution. Two years later, supply availability for these items was up 5 points for frequently demanded parts and 10 points on infrequently demanded parts. Inventory was simultaneously reduced by $600 million as the excess built up by forecasting was disposed of. The procurement workload was slashed—the number of purchase requests that required contracting action decreased by 35 percent and canceled purchase requests by 70 percent. In the second year of PNG implementation, DoD saved $400 million in working capital by better aligning purchases and sales.