Designing a Data-Driven Purchasing Process

Abstract

This study examines how a more data driven purchasing process can be designed at Jan Comstedt AB, a wholesale distributor operating in the Nordic region within the boating and sport fishing industry. The study was conducted in the context of a system transition from a legacy DOS-based system to Microsoft Dynamics 365 Finance and Operations. The transition was triggered by the acquisition by Alliance Marine in 2022, and acted for both a technical upgrade and a broader organizational transformation. The study has a mixed-method research design, combining semi structured interviews with employees across different departments with a quantitative analysis of 50 selected products from Comstedts products portfolio. The Gioia methodology was used to analyze the qualitative data and produced five aggregate dimensions covering experiences of the legacy system, the transition process, organizational change, data quality, and perceptions of the outcome. The quantitative section utilized ABC/XYZ classification, forecasting simulations using moving average and exponential smoothing evaluated through MAPE, and safety stock calculations based on cycle service levels for its analysis. The result from the study reveals that the technical foundation for a data-driven purchasing process exists within the new system of Dynamics 365. Theories such as ABC/XYZ classification and safety stock models are implementable with the existing data. However, the result of the forecast accuracy revealed that the data quality appeared to be a critical barrier. The result was severely limited by misplaced demand caused by stockouts, untagged one-off orders, and generalized lead times set in 30-day intervals. These issues resulted in MAPE values of approximately 95-99% and a near complete collapse of the XYZ classification, with 48 out of 50 products classified as Z-items. The qualitative findings revealed how the new system's capabilities are utilized and affected by the employee perceptions of the change. The results were not in line with existing change management theory, despite a top-down implementation with limited formal training, resistance was low. The results were influenced by the legacy system which was widely regarded as outdated. Nevertheless, uncertainty, selective trust in system outputs, and unrealized potential remain prevalent across the organization. The findings from the study concludes that a hybrid approach for purchasing should be applied. Using system outputs as a first signal which should be reversed and validated through qualitative judgement. Improving data quality is an essential rather than a suggestion. The purchasing function should also develop a more strategic role which requires both continued technical configuration and active organizational commitment.