经济
计量经济学
相(物质)
需求预测
环境科学
经济预测
时间序列
协整
指数平滑
系列(地层学)
作者
Robyn Goldsmith,Anna‐Lena Sachs,John Boylan
标识
DOI:10.1016/j.ijforecast.2026.09.001
摘要
Inventory planners make high-stakes, long-term decisions when a product reaches its end-of-life (EOL). The further a product is in the EOL phase, the smaller and more infrequent demand quantities become. Consequently, manufacturers often cease production when a part enters the EOL phase, meaning inventory decisions must be made for a long planning horizon, e.g., approximately 10 years for automotive spare parts. Accurate demand forecasts are fundamental for good inventory decisions but challenging due to declining sales whereby demand can become intermittent. This paper introduces a long-term stochastic forecasting model for stock keeping units (SKUs) in the EOL phase. Theoretical results into the bias and variance of model parameter estimates motivate an extension to a multiple series approach that incorporates demand observations from other series with the same decline rate. We find that the multiple series approach increases forecast accuracy, measured as mean absolute percentage error, and, while achieving a higher fill rate, incurs less leftover inventory. The increase in forecast accuracy is proportional to the number of series in the multiple series approach. Using 400 demand series collected at an automotive manufacturer, we show that our multiple series extension outperforms the benchmark method for forecasting demand in the EOL phase.
科研通智能强力驱动
Strongly Powered by AbleSci AI