What Soham demonstrated

Built an end-to-end demand forecasting analysis on 9,648 Adidas transactions: cleaned and resampled the sales history, explored regional and product patterns, decomposed the monthly series, and compared moving average, linear regression, and SARIMA models. The final SARIMA model achieved the best holdout accuracy (3.0% MAPE) and was translated into board-ready recommendations on regional inventory and channel strategy.

• Leakage-Free Time-Series Splitting
• Comparative Forecast Model Evaluation
• Seasonality and Trend Interpretation
• Business-Oriented Insight Synthesis