A Cloud-Native Framework for Cross-Industry Demand Forecasting: Transferring Retail Intelligence to Manufacturing with Empirical Validation 1 st Tingting Lin* SAP SE San Francisco, California, USA career.tingtinglin@gmail.com ORCID: 0009-0009-4556-5287 2 nd Sushma Kukkadapu Walmart (Sam’s Club) Bentonville, Arkansas, USA sushmakuk24@gmail.com ORCID: 0009-0007-6185-5271 3 rd Greeshma Suryadevara Walmart Bentonville, Arkansas, USA greeshmasuryadevara42@gmail.com Abstract—This digital transformation of manufacturing processes presents unique challenges in adapting proven retail forecasting methodologies to industrial settings. This paper introduces a novel cloud-native framework that bridges the gap between retail and manufacturing demand forecasting, leveraging advanced data analytics and scalable architecture. We present (1) an adaptive methodology for translating retail demand patterns to manufacturing contexts, (2) a cloud-native architecture supporting real-time data integration and scalable processing, and (3) empirical validation using public retail and manufacturing datasets. Our framework demonstrates significant improvements in forecast accuracy (15.0% reduction in Mean Absolute Percentage Error) and resource utilization (40% reduction in computing costs). Analysis across 50 manufacturing facilities shows that our retail-derived demand forecasting techniques effectively optimize manufacturing processes while maintaining 99.999% system availability. The framework achieves a 25% reduction in safety stock levels and 35% improvement in inventory turnover, providing practical insights for organizations pursuing digital transformation initiatives in manufacturing. Keywords: Cloud Computing, Demand Forecasting, Digital Transformation, Manufacturing Optimization, Retail Analytics, Scalable Architecture, Supply Chain Management I. INTRODUCTION The convergence of retail and manufacturing intelligence represents a critical frontier in industrial digital transformation. While retail sectors have developed sophisticated demand forecasting systems leveraging real-time data and cloud computing, manufacturing environments often lag in adopting these advanced predictive capabilities [4]. Recent advances in microservice architectures have begun to address this gap, enabling more flexible integration of manufacturing intelligence systems [1, 2]. This gap presents both a challenge and an opportunity for cross-industry knowledge transfer, particularly in the context of Industry 4.0 initiatives. The digital transformation of manufacturing processes represents a critical inflection point in industrial evolution, particularly as organizations seek to leverage advanced analytics and cloud computing capabilities. Traditional manufacturing planning systems operate within rigid frameworks, relying on historical data and static forecasting models. These approaches struggle with rapidly evolving supply chains and market demands, particularly during disruptions or unexpected consumer behavior shifts. The retail sector, however, has successfully implemented sophisticated systems leveraging real-time data streams, machine learning algorithms, and cloud processing. These systems effectively handle seasonal variations, promotional events, and demand shifts, making a compelling case for cross- industry knowledge transfer [2]. The success of modern industrial systems can be attributed to several key innovations, following strategic initiatives like Industrie 4.0 [3]. First, the integration of multiple data sources enables comprehensive demand predictions. Second, the adoption of cloud-native architectures facilitates scalable processing. Third, the implementation of machine learning algorithms enables continuous adaptation to changing market conditions. Translating these methodologies to manufacturing presents several challenges. First, fundamental differences exist in data structures—retail data centers on consumer transactions, while manufacturing involves complex interconnected variables including material availability, capacity constraints, and multi- tier dependencies. Manufacturing environments also present unique complexity challenges with longer lead times and intricate production dependencies. The forecasting system must account for equipment capacity, worker availability, material lead times, and quality control processes simultaneously. Finally, system integration poses significant challenges. Manufacturing facilities typically use established MES and ERP platforms with proprietary formats and protocols. Integrating new forecasting systems requires careful consideration of compatibility and interfaces while maintaining reliability during critical operations [4]. Our research examines the adaptation of retail forecasting methodologies to manufacturing environments through a systematic evaluation of cross-industry implementation strategies. We analyze public datasets from both retail and manufacturing sectors to demonstrate significant improvements in forecast accuracy and operational efficiency through our proposed framework. Building on established methodologies from Thomassey and Fiordaliso's hybrid forecasting system [5], our analysis shows that retail-derived forecasting techniques 2025 5th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA) 979-8-3315-0976-7/25/$31.00 ©2025 IEEE 1115 2025 5th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA) | 979-8-3315-0976-7/25/$31.00 ©2025 IEEE | DOI: 10.1109/AIITA65135.2025.11048053 Authorized licensed use limited to: Tingting Lin. Downloaded on July 07,2025 at 23:21:37 UTC from IEEE Xplore. Restrictions apply.