Background

Demand planning is one of the most valuable AI application areas in retail, but forecast accuracy alone does not create value: forecasts must be embedded in a planning process, presented to planners in a comprehensible form, and connected to a replenishment decision. Empirical work on retail forecasting repeatedly shows that planners override model recommendations they do not understand or trust, so that the design of the surrounding decision support system matters as much as the underlying model.

Modern forecasting libraries and pre-trained time-series models have made accurate baseline forecasting comparatively accessible, shifting the interesting research question from model engineering to requirements, integration, and use.

Object of investigation

Following a Design Science Research approach, this Master's thesis elicits requirements for an AI-supported demand planning decision support system, designs the artifact, and implements a prototype using established off-the-shelf forecasting components. The emphasis lies on the design of the human decision interface, the integration of the forecast into the replenishment decision, and the evaluation of the artifact with prospective users — not on developing new forecasting algorithms.

Possible research questions

• Which requirements do retail demand planners place on an AI-supported planning system?
• How should forecast uncertainty and forecast rationale be presented so that planners can judge when to intervene?
• How does the prototype perform in an evaluation with users, and what design principles can be derived?

Possible results

• A requirements catalog and a set of design principles for AI-supported demand planning.
• A working prototype demonstrated on a public retail dataset and evaluated with domain experts.

Starting literature

• Fildes, R., Ma, S. & Kolassa, S. (2022). Retail forecasting: Research and practice. International Journal of Forecasting, 38(4), 1283–1318. DOI 10.1016/j.ijforecast.2019.06.004.
• Makridakis, S., Spiliotis, E. & Assimakopoulos, V. (2022). The M5 accuracy competition: Results, findings, and conclusions. International Journal of Forecasting, 38(4), 1346–1364. DOI 10.1016/j.ijforecast.2021.11.013.
• Becker, J. & Winkelmann, A. (2019). Handelscontrolling: Optimale Informationsversorgung mit Kennzahlen, 4th ed. Springer Gabler.
• Peffers, K. et al. (2007). A Design Science Research Methodology for Information Systems Research. Journal of Management Information Systems, 24(3), 45–77. DOI 10.2753/MIS0742-1222240302.