Thesis/Capstone
Publication Date
Authored by
Bishwajit Kumar, Pablo Andres Barros Gomez
Abstract

An organization’s procurement process is pivotal for its success in a competitive market. The increased uncertainty and complexity of post-pandemic supply chains have made procurement a more valuable focus point among organizations and a differentiating factor to achieve a competitive advantage. The sponsor company of this study believes that the key to being competitive in today’s VUCA (volatile, uncertain, complex, and ambiguous) market relies on getting faster insights into the problem areas, having enhanced decision-making capabilities, and optimal exception management. For that reason, it seeks to understand if said competencies are encapsulated within a Procurement Control Tower’s value proposition. To meet our sponsor’s requirements, we divided our research into two components, a qualitative and a quantitative component. The first evaluates and defines the scope, value proposition, and deployment strategy of the Procurement Control Tower. The latter provides proof of concept by creating a working prototype of one of its use cases. The selected use case for the prototype is Spend Analytics, more specifically, the categorization and sub-categorization of the unclassified spend data for an assigned business unit. To create the prototype, this study compares multiple Machine Learning algorithms and selects Random Forest as the best-performing one in terms of accuracy. The algorithm’s predictive power is then enhanced by pre-processing the data with Natural Language Processing. The final model performs with 94% accuracy at a category level and 90% at a sub-category level. This study's primary finding, obtained through the categorization of approximately 250 million USD of unclassified spend data, is that implementing the Procurement Control Tower in the sponsor's business provides measurable value. For our assigned business unit, it creates renegotiation opportunities with suppliers, increases budgeting accuracy, and reduces the man-hours required. The final algorithm of the prototype has been presented to the sponsor company, which is currently deploying it for the assigned business unit. To scale up the benefits of the solution across the organization, the sponsor plans to deploy it for the remaining business units.

Attachment(s)