Time Series Foundation and Quantum Models for Demand Forecasting in Enterprise Supply Chain Planning

Enterprise demand forecasting in complex supply chains demands scalable, accurate, and computationally efficient solutions that integrate seamlessly into industrial planning systems. This work investigates the applicability of state-of-the-art time series foundation models (TSFMs) against classical statistical and machine learning baselines, assessing their zero-shot and fine-tuned inference capabilities across industry-standard benchmarks and real-world customer datasets. Key enterprise-relevant properties are systematically examined, including covariate selection, prediction interval estimation, and model robustness under missing data and temporal distribution shifts. Forecasting utility is further characterized across multilevel product, location, and category hierarchies within a top-down planning framework, demonstrating where TSFMs yield measurable accuracy gains in production SAP Integrated Business Planning (IBP)pipelines.

While TSFMs advance the classical frontier, demand forecasting in enterprise supply chain planning is constrained by the accuracy limits of classical models when faced with high-dimensional data and while quantum machine learning offers theoretical advantages yet empirical studies for this specific task are scarce. We present an empirical proof of concept integrating quantum forecasting models into SAP IBP, benchmarked systematically against established classical results. Acknowledging current quantum hardware constraints, we refrain from asserting quantum advantage and instead contribute a transparent, replicable benchmark positioning enterprise demand planning at the frontier of both classical AI and emerging quantum computing paradigms as algorithms and hardware continue to mature.