Forecasting Analysis of Cement Selling (Non-Bulk) Using The Method of Triple Exponential Smoothing (Case study: PT. Lafarge Holcim Cement Indonesia)

DOI: 10.31284/j.iptek.2020.v24i2 .311 PT. Lafarge Holcim Cement Indonesia is one of the cement companies focusing on construction. The aim of this study is to analyze the sales forecasting applied to know the next period using Triple Exponential Smoothing. The results showed that the values of MAD, MSE and MAPE for PCC's bag are 3980, 20925291 and 13%, respectively. Meanwhuile, the values for OPC's bag are 105, 19497 and 14%, respectively. Furthermore, the forcasting result of the next period of cement selling for PCC's and OPC's bag are 32498 and 792, respectively. Triple exponential smoothing method is a suitable method to forecast the case that has small bias, such as the case in PT. Lafarge Holcim Cement Indonesia.


INTRODUCTION
Cement is the primary raw material in building construction. To date, nothing has replaced the function of cement itself. Moreover, with the development of increasingly advanced technology, public facilities, infrastructure and housing increases in the number and in population. This has made the demand for cement also increase. The competition of cement business is also experiencing rapid development. Thus, the companies must have their respective strategies to maintain their company's continuity and success. PT. Lafarge Holcim Cement Indonesia (PT. LHCI), One of the cement companies in Aceh , produces two types of cement products, which are OPC (Ordinary Portland Cement) and PCC (Portland Composit Cement). Both have differences in the composition of raw materials based on the SNI standards.
Currently PT. LHCI still has difficulty in predicting future sales for increasing company profits. This is due to the company has not implemented an appropriate method of sales forecasting e-ISSN:2477-507X Vol.24 No.2, December 2020 76 . If the company takes the wrong action in predicting sales, the company might suffer losses. if the sales forecasting of the company is too high, the company might suffer losses due to the enormous production costs. Otherwise, if company management determines the sales forecasting too low. The company might also experience losses due to the inventory out of stock which not fulfilling consumer demand. Therefore, companies need to forecast their next selling using specific methods. This study aims to forecast cement selling using triple exponential smoothing method in PT. Lafarge Holcim Cement Indonesia.

Selling and Forecasting
Selling is buying and selling process of goods or services carried out in a place/market or online, using a legal payment instrument to make a profit. Forecasting is an objective calculation using past data to determine the future conditions. This forecasting has two methods, namely qualitative methods, and quantitative methods. the qualitative method is a method that simply analyzes the objective conditions. Meanwhile, quantitative methods analyze the patterns of product and sales development [1]. Quantitative forecasting have several methods, one of which is the triple exponential smoothing method. This triple exponential smoothing method is the analytical method for selling forecasting in this study.

Triple Exponential Smoothing Method
This method is proposed by Brown, using quadratic equations. This method is suitable for forecasting fluctuations. The forecasting procedure is as follows: [2] 1. Determning the value of S't with the following formula: For the first period, the value of S'1 cannot be found using the formula above but it can be determined freely. It is generally determined by the same value that occurred in the first period.
2. Determining the value of S "t using the formula: S" t = α S' t + (1 -α) S" t -1 In the first period, the value of S " t is usually determined as the value that occurred in the first period.
3. Determining the value of S "'t with the following formula: S"' t = α S" t + (1 -α) S"' t -1 For the first period, S" ' t is generally considered the same as the first period data.
4. Determining the a t value using the following formula: a t = 3 S' t -3 S" t -S"' t 5. Determining the bt value using the following formula: 6. Determining the value of ct using the following formula: . Producing a forecast equation as follows: F t + m = a t + b t m + ½ c t m² m is the forward period for how many future periods the forecast is carried out. a t , b t , c t are values calculated according to the formula above. There are several calculations commonly used to calculate the total forecast error. This calculation can be used to compare different forecasting models, as well as to monitor forecasts, and to ensure that the forecast properly function. The validation of forecasting methods, especially using the methods above, cannot be separated from the indicators in measuring forecasting accuracy. There are several indicators for measuring forecasting accuracy, but the most commonly used are the mean absolute deviation, mean squared error, and mean absolute percent error. [3] 1

. Mean Absolute Deviation (MAD)
Forecasting accuracy is high if MAD's values, mean absolute percentage error, and mean squared error are getting smaller. MAD is the absolute total value of the forecast error divided by the data. The formula for calculating MAD is as follows: The error value is calculated using the average squares difference between predicted and observed values. Mean Squared Error is also known as forecasting error. This forecasting error can also calculate the MAD value, which was discussed in the previous section. Forecast errors cannot be avoided in forecasting systems, but forecast errors must be appropriately managed. Management of forecast errors can be more effective if forecasters can take appropriate action regarding the forecast error. Various forecasting models provide different forecast values and different degrees of forecast error in a forecasting system. The average of square error strengthens the effect of large error rates but minimizes forecast error rates smaller than one unit.

Data Analysis
The data used for data analysis is the actual sales data obtained from PT. Lafarge Holcim Cement Indonesia from 2014 to 2016 for each type of cement. These data are then analyzed using the triple exponential smoothing formula, which is then tested for accuracy using Mean Absolute Deviation (MAD), Mean Squared Error and Mean Absolute Percent Error (MAPE). Next is the selling forecast for each type of cement in the following years. The last is the conclusions and suggestions on how to forecast cement selling using the method of triple exponential smoothing.

First Data Discussion
Before calculating the sales forecast, the first thing to do is to record the actual sales. In this study, sales data were taken from 2014 to 2016. Data on sales of non-bulk cement at PT Lafarge Holcim Cement Indonesia can be seen in Table 1

Discussion of Selling Forecasting for OPC Cement
The calculation of forecasting non-bulk cement selling using the triple exponential method for OPC cement type can be seen in table 3.