A beverage manufacturer needed to accurately forecast daily empty bottle returns for 10 SKUs to optimize production planning. They developed predictive models using 2 years of returns data and techniques like ARIMA, Holt Winters, and year-on-year growth. The models achieved 75% accuracy on average for May-June 2011 forecasts, and 92% accuracy for the highest-return SKU specifically. This significantly improved upon existing forecasting methods.