Integrated Time Series Analysis for Long Term Demand Planning and Capacity Expansion

Within the dynamic landscape of the fast-moving consumer goods (FMCG) industry, a food company encountered challenges associated with fluctuating demand forecasting, resulting in inefficiencies in determining production volumes for various product formats. The paper outlined a comprehensive methodology applied to tackle these challenges, it incorporated the use of exponential smoothing (specifically Holt-Winters' additive with dampening) and autoregressive integrated moving average (ARIMA) models, alongside a combination method to enhance forecast reliability. Forecasting models were evaluated through residual diagnostics and performance metrics such as root mean squared error (RMSE) and mean absolute percentage error (MAPE). The report delved into prediction intervals and scenario-based forecasting, and it provided a holistic perspective on managing forecast uncertainty. To sustain enhanced forecasting, tracking signals were employed to detect persistent bias. In production planning, this paper assessed line utilization based on forecasted data. Results indicate that the combination of ARIMA yielded the best performance with a MAPE of 3.75%. Additionally, over three years, line utilization averaged 94% in normal states, emphasizing the effectiveness of the forecasting methods in production planning. Furthermore, the paper recommends the ongoing use of the ARIMA combination method and emphasizes the incorporation of tracking signals for model validation to enhance accuracy and reliability. Looking ahead, the paper suggests considering an expansion in the five-year plan, aligning production capabilities with forecasted demand to further optimize operational efficiency in the FMCG industry. The proposed model sets the stage for sustained efficiency and strategic growth in the ever-evolving FMCG landscape.