Quantitative Demand Forecasting of Spare Parts in the Aviation Industry
by HAVADER Editör Ekibi
One of the most common reasons an aircraft sits grounded is simply that a needed spare part isn't in stock. For airline maintenance and repair organizations (MROs), this is both costly and operationally risky: stock too many parts and you tie up capital and warehouse space; stock too few and you can leave an aircraft stuck in a hangar for days. It's the same dilemma anyone has faced waiting for a car part to arrive — just scaled up to an industry where delays can ground a multi-million-dollar aircraft.
The purpose of this study was to determine which numerical forecasting method most accurately predicts future failure counts for the 10 parts a large airline MRO replaces most often. Using monthly failure data from 2021-2022, the researchers pitted four different statistical forecasting methods against each other and then checked their predictions against actual 2023 outcomes.
Each method's accuracy was measured using Mean Absolute Error, Mean Square Error, and Mean Absolute Percentage Error. The result was strikingly clear: the Simple Exponential Smoothing method produced the most accurate forecasts, while a considerably more complex method, Holt-Winters, performed worst on this particular dataset — a useful reminder that "more sophisticated" doesn't automatically mean "more accurate" in aviation forecasting.
What makes this study valuable is that it demonstrates a concrete, measurable benefit to using quantitative forecasting for defective-parts planning. When an MRO firm adopts this approach, it can avoid unnecessary inventory costs while also reducing the risk of grounding an aircraft for days because a critical part wasn't on the shelf. The everyday parallel is familiar to anyone who has dealt with a car mechanic predicting which spare parts to keep in stock and how often — except here the stakes are much higher, since delays ripple directly into passenger flights.
The study also opens a door for future research: whether machine learning algorithms could push forecasting accuracy even further is likely to remain an active question in aviation maintenance for years to come.
← Blog
The purpose of this study was to determine which numerical forecasting method most accurately predicts future failure counts for the 10 parts a large airline MRO replaces most often. Using monthly failure data from 2021-2022, the researchers pitted four different statistical forecasting methods against each other and then checked their predictions against actual 2023 outcomes.
Each method's accuracy was measured using Mean Absolute Error, Mean Square Error, and Mean Absolute Percentage Error. The result was strikingly clear: the Simple Exponential Smoothing method produced the most accurate forecasts, while a considerably more complex method, Holt-Winters, performed worst on this particular dataset — a useful reminder that "more sophisticated" doesn't automatically mean "more accurate" in aviation forecasting.
What makes this study valuable is that it demonstrates a concrete, measurable benefit to using quantitative forecasting for defective-parts planning. When an MRO firm adopts this approach, it can avoid unnecessary inventory costs while also reducing the risk of grounding an aircraft for days because a critical part wasn't on the shelf. The everyday parallel is familiar to anyone who has dealt with a car mechanic predicting which spare parts to keep in stock and how often — except here the stakes are much higher, since delays ripple directly into passenger flights.
The study also opens a door for future research: whether machine learning algorithms could push forecasting accuracy even further is likely to remain an active question in aviation maintenance for years to come.