The AF-Filières (Supply Chain Flow Analysis) methodology, implemented in the SOCLE project, aims to reconstruct all the flows of a supply chain by aggregating, structuring and analysing all the available information. It consists in gathering all the relevant data about a given sector into a single, hierarchical model that links every piece of information together. This systematic cross-checking makes two essential operations possible: 1. checking the internal consistency of the data and, where necessary, reconciling it; 2. determining missing values, either precisely or as intervals when some indeterminacy remains.
The Sankey diagrams presented in the "Method and sources" section illustrate each step of this process, relying directly on the data actually used in the SOCLE model.
(Associated diagram: Sources)
This first phase starts with an inventory of all the relevant public and sector-specific sources describing the supply chain (available years, nature of the data: direct values, min/max, coefficients, etc.). The associated diagram shows, for each documented flow, the sources used and their characteristics, giving an immediate reading of the well-documented areas and of those where information is scarcer. The diagram also makes it possible to display the type of data collected in the legend: direct value, or allocation and yield coefficients, which express a relationship between two flows rather than an absolute value. The raw data from the source can also be consulted in the corresponding sheet of the model's Excel file (the sheet name is in the tooltip of each flow).
Because the collected data generally does not cover the whole supply chain, this step also includes structural completion: additional flows, not yet documented, are added to the model to ensure the continuity of the processing chains.
At the end of this phase, the whole knowledge perimeter is mapped: the flows for which data exists and those that will have to be determined later.
(Associated diagram: Method)
Based on this complete structure, the flows for which reliable data exists are consolidated. When several sources converge, the collected values are retained as they are. When inconsistencies are detected (contradictory material balances, incompatible yields, etc.), a statistical discrepancy flow may be added to the model.
Once the model has been made consistent, the missing flows are determined using material balances, structural constraints and relationships between flows. Some flows thus obtain a precise value ("determined"); others can only be defined through a min–max range when a lack of information remains. In the wine supply chain, the fermentation losses (CO₂) are thus determined by material balance on the winemaking node, and out-of-home consumption by balance on the marketed wine.
(Associated diagram: Confidence level)
The last diagram offers an overall view of the results, enriched with a global reliability indicator for each flow. This reliability is based on several criteria: quality and precision of the sources, closeness of the collected data to the project's perimeter, magnitude of the statistical discrepancies, width of the indeterminacy intervals, etc.
How to read the reliability levels:
This diagram makes it possible to identify at a glance the segments of the supply chain that are well documented (harvest, trade, stocks) and those that remain more exploratory (out-of-home consumption, the fine breakdown of co-product outlets).
