In managing inventories of spare parts for military equipment, it’s important to relate dollars invested in spares to readiness of the equipment. Hundreds of millions of dollars and the effectiveness and efficiency of military operations worldwide rely on this concept of readiness, or the “availability” of an aircraft, fleet of ships, or other warfare machinery to be ready for use and not inoperable for lack of a part. To predict this availability, agencies use readiness-based sparing (RBS) models, software tools that seek to show the tradeoff between spares investment and availability at various levels.
RBS Assumptions and Math
RBS models started as sophisticated mathematical tools using complex algorithms and have transformed over time to have better user interfaces and analytics capabilities. While many versions exist, the underlying mathematical assumptions and algorithms are substantially similar across all of them; they all use statistical distributions and probabilities to model uncertainty of demand or lead times. These probabilities are used to compute expected backorders and assess how many weapon systems are likely to be down at any given time. But these probabilities throw away much of the useful information in the raw data and distort the true picture of the uncertainty. So what happens to the pretty predictions when the equations don’t truly describe or fit the data at hand?
Is RBS Math Realistic?
RBS vendors, like other software vendors, emphasize the features and ease of use of their software but leave out the assumptions and limitations inherent in the model. They’ll typically tell prospective buyers that most clients’ data aligns with their math assumptions, but in reality, one size does not fit all. Building inputs for RBS requires forecasting mean and variance of items’ demands—this process throws away much of the useful information in the raw data. Beyond that, the mean and variance are unstable over time. This consequentially leads to inventory or investment levels and weapon system availability that are also unrealistic. But just how skewed from reality will my results be?
We were curious too, so we tested RBS modeling against recent historical data in an inventory simulation using LMI’s Financial and Inventory Simulation Model (FINISIM) tool, a tool that provides independent assessments of RBS and other similar tools without assuming mathematical equations. The simulation creates potential future streams of demand based on historical demand transactions—maintenance or repair activities’ requests. It then provides an objective performance assessment of how RBS modeling stacks up against those potential future streams of demand that reflect history, demonstrating just how far expectation is from reality.
What Did We Find?
To test the assumed statistical distribution for demands, we let RBS compute stock levels using perfect advance knowledge of the mean and variance of demands over the history, something you’d never have in reality but eliminating mean and variance as sources of error.
RBS predictions are not close enough to reality, and FINISIM has proven that. So if you’re not asking, “What do I really get from readiness based sparing with my data?” you should be!
“LMI gives me the freedom to follow my vision in solving some of the hardest problems in government, starting with first principles, and avoiding formulaic approaches. LMI has supported me in building an intelligent, creative, and resourceful team of people who help each other succeed. Being part of this team is a privilege!”