M factor and it’s generation process in demand forecast : risk-informed demand forecasting : M factor for resilient inventory

Abstract

Accurately forecasting spare-parts demand is pivotal to airline inventory efficiency, yet conventional methods often overlook the gap between design-stage reliability and in-service performance. This study proposes the M factor, defined as the ratio of supplier-reported MTBF to the airline’s field MTBF and adjusted by a user-defined RISK term, as a compact indicator of that gap. After cleansing nearly 100 000 removal records for Boeing components, Time-Between-Failures (TBF) were extracted and fitted with Weibull distributions; median TBF was adopted to mitigate heavy-tailed noise. An experimental data set containing M, TBF, unscheduled removals and model-series tags was then fed into a random-forest regression pipeline implemented in Python. The resulting demand-forecast model achieved a five-fold cross-validation RMSE of 1.50 ± 0.27 and a held-out test RMSE of 0.99, demonstrating that incorporating the M factor significantly improves prediction accuracy over baseline features alone. Scenario analysis further shows that overestimating MTBF (M > 1) risks under-stocking and downtime, whereas underestimating MTBF (M < 1) inflates holding costs. The proposed framework offers both a theoretical bridge that linking reliability statistics with inventory theory, and a practical, reproducible Python tool-chain; future work will couple advanced machine-learning models with dynamic M-factor trends to enable fully adaptive spare-parts strategies.