Supplier Risk Assessment Under Uncertainty: Developing a Structured Evaluation Approach

The analysis of supplier risk has become more complicated due to increased complexity of the supply chain, increased level of disruption and uncertainty. Currently available frameworks mostly operate with data on quality, costs and delivery. They do not produce a complete picture of the risks involved. In addition to that, one of the problems faced by companies in assessing their risks is obtaining the relevant information and incorporating it into the assessment. This thesis explores an alternative methodology for conducting data-driven supplier risk assessment in the described situation. This includes analysis of combining internal quality data, logistics data and selected external data sources and determining whether logistics and external data bring value beyond the internal quality data. In order to do that, the thesis uses a case of Altromercato organization. The data used include various data sources related to operations, quality and logistics metrics extracted from the company’s shipment records as well as country-level risk indicators from the external data sources. The work follows the CRISP-DM framework and involves stages of data preparation, modeling and evaluation. The framework evaluates the value of a particular data source through the modeling scenarios. The statistical screening is used to determine the predictors associated with shipment results. The regression models and Isolation Forest algorithm are applied to explore potential patterns that indicate risk cases. The main finding is that the value of a data source does not depend only on its relevance, but also on its availability and consistency. Instead of considering data constraints as peripheral aspects, this thesis focuses on them as core elements of a supplier risk frame-work. The key contribution of this work is the proposed methodology for organizing and combining various data sources of the supplier risk in the real world.