Stochastic S&OP, taking uncertainties into account in the Sales and Operations Planning process

The current literature, although relatively limited, has focused on the development, management, and coordination of the S&OP process. After more than 30 years of existence, the S&OP process seems to be reaching an inflection point due to market changes, technological advancements, and the proliferation of data. This turning point is driven by the need for more effective risk management to create more resilient supply chains, especially in the face of factors such as frequent global crises (environmental, energy, geopolitical...) that have significantly disrupted the flow of products and raw materials. An increase in the frequency of short-term disruptions (demand spikes, raw material shortages) with strategic impacts is anticipated.
The integration of S&OP with operational planning and execution thus becomes crucial. In this context, numerous projects are underway to rethink forecasting and planning in supply chains, with a particular emphasis on leveraging all available data, thereby offering unprecedented predictive capabilities. Here, we examine the current state of the S&OP process and the associated demand forecasting approaches. In order to mitigate risks and improve decision-making, several studies have explored forecasting approaches that combine statistics and machine learning (AI). Research has examined deep learning-based approaches to predict sales. Other studies have investigated the usefulness of data and its exploitation to improve forecasts in supply chains. Additionally, other research has developed discrete event simulation models to identify risks and associated scenarios in supply chains.
This research project aims to develop effective forecasting models that mathematically integrate uncertainty. The goal is to identify the types of risks and opportunities specific to the automotive industry, capture relevant data, and combine them in a way that minimizes overall uncertainty. The thesis aims to improve the quality of S&OP predictions by developing advanced models that integrate uncertainty, identify risks and opportunities specific to the automotive industry, and propose effective strategies to reconcile commercial and industrial perspectives.