Bridging the gap between future uncertainties and demand forecast

Abstract

In supply chain bottlenecks start from wrong demand forecasts that later affect the whole production process, it is therefore urgent that managers know the importance of understanding how to deal with demand fluctuations. In general, the main issues addressed are: the level of detail of the aggregation of demand, the categorization of demand in different classes and the association of the different demand categories to their specific statistical model. Nowadays, demand planners lack a general procedure to follow while making forecast for future previsions. The main goal of this research is two-folded: the creation of a universal forecasting process that can be both accessible for a managerial perspective and innovative compared to the theoretical works that have been performed until now. In this way managers would not see the forecasting process as a “black box” anymore because one of the main aim is to give accessible and adaptable explications on how different mathematical models can be used according to the market at stake.

The analysis is divided into three main parts: the development of an up-to-date list of forecasting models for future demand is conducted in parallel to the construction of a new market’s demand classification. The results coming from these two first analyses are then linked together through the creation of a decision tree that managers would use to associate each market’s demand category to its specific set of statistical models. Therefore, the main contribution of this research is the development of precise strategies that managers can adopt to improve operational performance when demand is intermitted.
The main sources of information of this study are both the company taken under analysis and the past literature review. Concerning this latter, a lack of a universal view on intermitted demand is highlighted: the issue of classification is still in its infancy phase, the optimization of statistical forecasting methods is still an open topic for researchers and in addition to this when considering demand classification method, the past literature is hardly ever taking into account the different nature of items within an industrial context.

Accessibility, clarity and completeness are the three main requirements of the support looked for during this research. A prescriptive decision making tree is hence obtained as the final deliverable of this study, in order for the demand planners to take the best decisions during the forecasting journey. The exceptional advantage of this support is that compared to statistical tools where classification’s rules are represented by a group of numbers and formulas, decision trees give a symbolical reproduction of the decision making journey.

The final results highlight that the final model developed in this research performs better in the 59% of the cases under analysis. This research therefore embodies five main contributions to the past literature research: (1) the knowledge theory concerning the forecasting process is brought close to the everyday managerial practice (2) this work strongly contributes to the categorization of the market demand not only from a theoretical perspective but also from a practical one by implementing it in a real industrial case (3) thanks to the clustering of the demand, this research takes into account the different nature of items within the industrial context, (4) forecasting methods are not considered as a “black box” anymore by demand planners and (5) this research creates a holistic support that can help managers during their entire forecasting journey. Finally, conclusive recommendations are given both for future lines of research and for the company at stake.