Identifying contributors to supply chain outcomes in a multi-echelon setting: a decentralised approach

TL;DR

This paper tackles the problem of attributing variations in supply chain outcomes to individual actors in a multi-echelon setting under strict data privacy. It introduces a decentralised explainable AI framework that uses neural network ensembles to estimate total uncertainty as a proxy for each actor's contribution, eliminating the need for sharing private process data and enabling privacy-preserving attribution. Empirical validation on real multi-stage manufacturing data shows performance comparable to centralised SHAP, with additional tests on tabular datasets demonstrating generalisability. The approach enables practical applications such as federated learning participant selection and autonomous SCM agents, offering a privacy-respecting, scalable alternative for identifying sources of quality and delivery variance in complex supply chains.

Abstract

Organisations often struggle to identify the causes of change in metrics such as product quality and delivery duration. This task becomes increasingly challenging when the cause lies outside of company borders in multi-echelon supply chains that are only partially observable. Although traditional supply chain management has advocated for data sharing to gain better insights, this does not take place in practice due to data privacy concerns. We propose the use of explainable artificial intelligence for decentralised computing of estimated contributions to a metric of interest in a multi-stage production process. This approach mitigates the need to convince supply chain actors to share data, as all computations occur in a decentralised manner. Our method is empirically validated using data collected from a real multi-stage manufacturing process. The results demonstrate the effectiveness of our approach in detecting the source of quality variations compared to a centralised approach using Shapley additive explanations.

Figures (12)

• Figure 1: A centralised contribution estimation setting in which the machine learning (ML) model has access to data from multiple companies.
• Figure 2: A decentralised contribution estimation setting in which the data of each company is not shared with others.
• Figure 3: Multi-stage process flow of production system with measurement aggregation into a quality score
• Figure 4: Feature values for Company A with datapoints for measuring errors removed
• Figure 5: Measurements with datapoints for measuring errors removed (excl. #4 due to >90% of data missing)