Systemic forecast within the global supply chain

Multiple-weighted moving averages, linear regression, exponential smoothing, statistical inference, stochastic models... these methods, widely represented in forecasting software packages, have proved their worth. However, globalization is constantly making it more complex to understand the parameters and variables involved in our forecasts, increasing uncertainty and making our confidence intervals tingle.

We can simplify the complicated, but not the complex; expert, "neural" and learning systems are a good response to this increase in complexity, but we still need to master their inputs and select the most representative ones.

Not all managers and operational forecasters are experts in statistics. The former tend to trust blindly, or reject uncompromisingly, scientifically calculated results. The latter, in their routine phase, end up losing some of the sense of their multi-parameterization and their capacity for analysis. Quite often, cross-functional strategic or tactical forecasting/planning meetings (PIC, BP...) are reduced to clashes of chapels where, for lack of demonstrability, after long and heated exchanges, the figures adopted are those of who speaks loudest.

Don't we end up forgetting to take the step back necessary to manage complexity? Don't we need to take off our glasses from time to time and look at the horizon?

In fact, most of us agree that, given the difficulty of forecasting, the use of two methods and/or scenarios is good practice.

The aim of this article is to propose, as a complement to more traditional methods, approaches and concepts that provide a different perspective in support of forecast estimation, across the board, without frustration, with a global rationale and ease of sharing, monitoring and steering. These proposals are aimed at forecasting sales, purchases of "independent" or common products, material, human and budgetary resources, for any type of product with a significant life cycle (two years minimum).