IMPROVING SUBSIDIZED FERTILIZER DISTRIBUTION PLANNING ACCURACY USING TIME SERIES FORECASTING AT PT PUPUK KUJANG CIKAMPEK

• Main Author: Harianto, Freddy
• Format: Theses
• Language: Indonesia
• Subjects: Manajemen umum
• Online Access: https://digilib.itb.ac.id/gdl/view/69976
• Institution: Institut Teknologi Bandung

PT Pupuk Kujang is a subsidiary of PT Pupuk Indonesia. PT Pupuk Kujang has 2 urea plants, namely K1A and K1B plants, each year producing relatively stable amounts of urea fertilizer in the range of 990,000 tons. This production is divided to meet the needs of subsidized fertilizers, non-subsidized fertilizers, and as raw material for other fertilizers. Through PT Pupuk Indonesia, PT Pupuk Kujang received an assignment from the government to meet the need for subsidized fertilizer in the provinces of Banten, DKI Jakarta, West Java and parts of Central Java. Production and availability of stock for subsidized urea fertilizer must be prioritized over non-subsidized urea fertilizer, so that the production and sale of non-subsidized fertilizer can be carried out if the calculation of the need for subsidized fertilizer can be fulfilled. If the demand for subsidized fertilizer is high but the realization is much lower than planned, there will be an excess stock of subsidized fertilizer and it cannot be sold as non-subsidized fertilizer. As a precautionary measure, potential excess stock will be produced and sold as non-subsidized fertilizer. However, sales to the non-subsidized sector will be less effective if carried out in a narrow time given the limited time and market absorption. So it is necessary to plan the distribution of subsidized fertilizers well to reduce the difference with the actual distribution. Based on the root cause analysis carried out using the Current Reality Tree, there are several results of the analysis that influence the discrepancy between distribution planning and its actual distribution. However, what can be controlled and is still within the scope of research is the lack of accurate forecasting methods. To obtain an appropriate forecasting method, several time series forecasting models are used including: Simple Moving Average, Weighted Moving Average, Exponential Smoothing, Exponential Smoothing with Trends, Linear Regression Analysis, Decomposition of a Time Series, and Decomposition Using Least Square Regression. The forecasting model with the lowest forecasting error rate of MAPE (Mean Absolute Percent Error) will be proposed to be used as a forecasting model that can be used by company. Based on calculations, for the West Java province the lowest MAPE is produced by the Decomposition Using Least Squares Regression model with a period of 12 months. The forecasting model and time period cannot be applied to all regencies in West Java province, only 7 regencies have the lowest MAPE in this model and period. While the remaining 11 regencies have the lowest MAPE in the forecasting model as follows: 8 regencies in the Decomposition of a Time Series model with a period of 12 months, 2 regencies in the Decomposition of a Time Series model with a period of 21 months, and 1 regency in the Decomposition Using Least Squares Regression model with a time period of 21 months.