The problem is the difficulty of forecasting lumpy demand. There are two broad approaches to spare parts forecasting: the first is based on the operational experience of an enterprise, and the second on the application of forecasting techniques. Only 10% of companies use forecasting models.
Airline operators usually base predictions on their operational experience, on annual budgets, and information from aircraft manufacturers that prepare lists of recommended spare parts. The aim is to evaluate and compare. When new types of aircraft are introduced, the airframe and engine manufacturers normally provide a list of recommended spare parts, which is based on the projected annual flying hours of the new aircraft. The original equipment manufacturers also provide overhaul manuals for aircraft components aimed at supporting assessment of required replacements, i.e. data on the operational life of components. Consequently, the forecast of spare part inventories is usually based on past usage patterns and the experience of company personnel.
Demand forecasting (in the mathematical sense of forecasting) for lumpy items is a complex problem. Previous studies are not very accurate. The very latest work by Regattieri, Gamberi, Gamberini and Manzini, using 6 years of data for an A320 fleet, found the best forecasting approaches to be weighted moving averages (which is also easy to do), the Croston method (exponential smoothing), and exponentially weighted moving average models. All do-able in a spreadsheet or a database, although no doubt some IT salesman can make a strong case for spending millions of dollars on their software.