<![CDATA[Inventory management at Micro, Small, and Medium Enterprises in traditional markets remains heavily reliant on intuition, posing high risks of overstocking or stockouts. This study develops a data-driven inventory requirement forecasting system for Toko Mandiri by integrating the Demand Response–Autoregressive Moving Average (DR-ARMA) method into a desktop application using the Rapid Application Development (RAD) approach. One year of historical daily sales data from four fermented products, with sparsity levels ranging from 38% to 51%, was utilized for training and testing. The dataset was partitioned into training (60%), validation (20%), and testing (20%). The DR-ARMA model's performance was evaluated quantitatively, yielding average values of 1.632 units for RMSE, 1.012 units for MAE, and 29.89% for MAPE, demonstrating reliability on fluctuating and sparse data. System usability evaluation involved three respondents across four task scenarios. Results indicate significant improvements in operational efficiency: the time required to determine inventory requirements was reduced from 30 minutes, using manual intuition-based methods involving physical stock checks across different locations, to just 1 minute, based on direct stopwatch measurements for all scenarios. This represents a 96.7% reduction in processing time. Interaction steps were streamlined from an unstructured process to only 6–7 clicks. The system's effectiveness reached an 83.33% task success rate among non-technical users. Integrating DR-ARMA into a practical application effectively transforms inventory decision-making from intuition-based to data-driven, potentially reducing operational risks.]]>
Copyrights © 2026
