Forecasting mail flow: A hierarchical approach for enhanced societal wellbeing

Abstract

Forecasting for Social Good has gained considerable attention for its impact on individuals, businesses, and society. This research introduces an integrated hierarchical forecasting-based decision-making approach for mail flow in a major postal organisation, presenting new social performance indicators. These indicators, including the discharge level, discharge rate, and overload rate, guide decision makers toward consistent workload planning, bridging a literature gap concerning forecast utility measures. The study evaluates three forecasting methods—exponential smoothing with error, trend, and seasonality (ETS), the autoregressive integrated moving average (ARIMA), and the light gradient boosting machine (LightGBM)—in terms of forecast accuracy and social measures, comparing them to the organisation’s current method. The empirical results confirm that the proposed approach is more accurate than the current method. Moreover, while ETS shows the highest forecast accuracy, LightGBM outperforms all methods in social measures. This indicates that a highly accurate forecasting method does not always enhance social performance, challenging traditional views on forecasting evaluation.