Predictive Lead-Time Analysis and Factory Reallocation for Nassau Candy Distributor

Factory allocation is an important supply-chain decision because manufacturing location can influence shipping distance, delivery time, workload concentration, and profitability. This study analyzes the US Candy Distributor dataset and develops an exploratory framework for evaluating shipping performance and potential factory reallocation. The study combines exploratory data analysis (EDA), feature engineering, geographic distance calculation, supervised regression, K-Means clustering, counterfactual simulation, and mixed-integer optimization. The dataset contains 10,194 historical order records covering 15 products and five factories. The analysis identifies total sales of $141,783.63, gross profit of $93,442.80, and an average recorded lead time of approximately 1,320.84 days. A major data-quality issue is that order dates occur in 2024–2025 while shipment dates extend into 2026–2030, producing recorded lead times between 904 and 1,642 days. This anomaly is retained rather than artificially corrected. Three regression models were evaluated, with Linear Regression reported as the best non-dummy model (MAE = 213.35 days, RMSE = 264.87 days, R² = -0.000537). The near-zero explanatory power indicates that the available transactional variables are insufficient for reliable operational lead-time prediction. Sixty counterfactual product-factory scenarios were evaluated, but no scenario passed the project's complete recommendation gate. The study therefore treats the optimization and reallocation outputs as exploratory decision-support results rather than operational recommendations.