Demand Forecasting and Production Planning

ScienceAsia 27 (2001) : 271-278

Demand Forecasting and Production Planning for Highly Seasonal Demand Situations: Case Study of a Pressure Container Factory
Pisal Yenradeea,*, Anulark Pinnoib and Amnaj Charoenthavornyingb a Industrial Engineering Program, Sirindhorn International Institute of Technology, Thammasat University, Patumtani 12121, Thailand. b Industrial Systems Engineering Program, School of Advanced Technologies, Asian Institute of Technology, P.O. Box 4, Klong Luang, Patumtani 12120, Thailand. * Corresponding author, E-mail: [e-mail]

Received 24 May 2001 Accepted 27 Jul 2001

ABSTRACT This paper addresses demand forecasting and production planning for a pressure container factory in Thailand, where the demand patterns of individual product groups are highly seasonal. Three forecasting models, namely, Winter’s, decomposition, and Auto-Regressive Integrated Moving Average (ARIMA), are applied to forecast the product demands. The results are compared with those obtained by subjective and intuitive judgements (which is the current practice). It is found that the decomposition and ARIMA models provide lower forecast errors in all product groups. As a result, the safety stock calculated based on the errors of these two models is considerably less than that of the current practice. The forecasted demand and safety stock are subsequently used as inputs to determine the production plan that minimizes the total overtime and inventory holding costs based on a fixed workforce level and an available overtime. The production planning problem is formulated as a linear programming model whose decision variables include production quantities, inventory levels, and overtime requirements. The results reveal that the total costs could be reduced by 13.2% when appropriate forecasting models are applied in place of the current practice. KEYWORDS: demand forecasting, highly seasonal demand, ARIMA method, production planning, linear programming,

References: 1. Nahmias S (1993) Production and Operations Analysis, 2nd ed, Irwin, New York. 2. Vandaele W (1983) Applied Time Series and Box-Jenkins Models, Academic Press, New York. 3. Winters PR (1960) Forecasting Sales by Exponentially Weighted Moving Average. Management Science 6(4), 324-42. 4. Box GE and Jenkins GM (1970) Time Series Analysis, Forecasting, and Control, Holden-Day, San Francisco. 5. Makridakis S Wheelwright SC and McGee VE (1983) Forecasting Methods and Applications, 2nd ed, John Wiley & Sons, New York. 6. Johnson LA and Montgomery DC (1974) Operations Research in Production Planning, Scheduling, and Inventory Control, John Wiley & Sons, New York. 7. Bullington P McClain J and Thomas J (1983) Mathematical Programming Approaches to Capacity Constrained MRP Systems: Review, Formulation, and Problem Reduction. Management Science 29(10). 8. Gabbay H (1979) Multi-Stage Production Planning. Management Science 25(11), 1138-48. 9. Zahorik A Thomas J and Trigeiro W (1984) Network Programming Models for Production Scheduling in MultiStage, Multi-Item Capacitated Systems. Management Science 30(3), 308-25. 10.Lanzanuer V (1970) Production and Employment Scheduling in Multi-Stage Production Systems. Naval Research Logistics Quarterly 17(2), 193-8. 11.Schwarz LB (ed) (1981) Multi-level Production and Inventory Control Systems: Theory and Practice, North-Holland, New York. 12.Tersine RJ (1994) Principles of Inventory and Materials Management, 4th ed, Prentice Hall, New Jersey. that the optimal inventory holding cost and overtime cost in the production planning model based on the recommended forecasting models are almost equal which indicates that the model can efficiently achieve a tradeoff between both costs. Normally, the optimal decisions in the first planning period will be implemented. After the first period has passed, the new forecasts will be determined, and the model parameters will be updated. The updated model is solved again to determine the optimal decisions in the current period. This is called a rolling horizon concept. However, the details and results of this step are not shown in this paper. DISCUSSION AND CONCLUSION The ARIMA model provides more reliable demand forecasts but it is more complicated to apply than the decomposition model. Therefore the ARIMA model should be used only when the decomposition model is inadequate. When compared against those of the current practice of the company, the errors of our selected models are considerably lower. This situation can lead to substantial reductions in safety stocks. Consequently, the lower safety stocks result in decreased inventory holding and overtime costs. The results of the production planning model are of great value to the company since the model can determine the optimal overtime work, production quantities, and inventory levels that yield the optimal total overtime and holding costs. The production planning method is more suitable than the existing one that does not consider any cost factors. Moreover, it has been proven that an application of appropriate forecasting techniques can reduce total inventory holding and overtime costs significantly. In conclusion, this paper demonstrates that an improvement in demand forecasting and production planning can be achieved by replacing subjective and intuitive judgments by the systematic methods.