Multi-level approaches to demand management in complex environments: An analytical model

Recent studies have shown that as demand becomes irregular and complex (i.e., lumpy), a possible approach for managing such uncertainty is to collect information directly from customers. This implies that the sales units have to move closer to customers, analyse their likely requirements, and collect quantitative and structured data as well as qualitative and subjective insights. However, as integration with individual customers increases and data collection capabilities improve the organisational configuration of most companies becomes ever more complex and the aggregation of forecasts more difficult. This paper discusses two approaches to managing demand uncertainty in complex environments. In the first (termed decentralised order overplanning), sales units are responsible for forecasting the demand of each customer and defining requirements. In the second (termed centralised order overplanning), forecasts provided by sales units are aggregated and further elaborated by manufacturing to define item requirements. By means of an analytical model (which describes the forecasting and planning process as a Bayesian–Markovian process), we show that the centralised method out-performs the decentralised approach by virtue of the ability to exploit the additional information provided by commonalities between customers' requests. However, this advantage has to be balanced against organisational costs. Since the centralised method splits responsibilities for forecasting and slack control between sales and manufacturing units, major conflicts are likely to arise, the focus and commitment on forecasting accuracy may be compromised, and information may be lost when individual forecasts are sent to the manufacturing unit.