|Sintia Putri Pradita||Department of Agro-Industrial Technology, Faculty of Agro-Industry, Kasetsart University, 50 Ngam Wong Wan Rd. Lad Yao, Chatuchak, Bangkok 10900, Thailand|
|Pornthipa Ongkunaruk||Department of Agro-Industrial Technology, Faculty of Agro-Industry, Kasetsart University, 50 Ngam Wong Wan Rd. Lad Yao, Chatuchak, Bangkok 10900, Thailand|
|Thaweephan Duke Leingpibul||Haworth College of Business, Western Michigan University, 1903 Western Michigan Avenue Kalamazoo, MI 49008, USA|
The demand for reefer containers in Indonesia has been increasing due to both global and regional trade growth; however, logistics providers are still struggling with several related challenges, including a container shortage problem, which is due to ineffective forecasting practices. This study aimed to improve the accuracy of reefer container demand forecasting by introducing an intervention forecasting approach. This approach will help address the demand planning issue of reefer circulation. The intervention forecasting approach combines human insights from the qualitative approach with the mathematical precision of the quantitative approach in iterative sequences. This field study was conducted with an Indonesian third party logistic company in Eastern Indonesia. The training data set was analyzed to provide a pattern of demand as well as some initial forecasting parameters (such as trend and seasonal index). Then, an expert helped identify irregular demand points. The demand data was then adjusted by a sales and marketing manager according to related factors such as natural disasters, oil price increase, promotions. The selected models were then further verified using a testing dataset, and the forecast errors from various models using the raw and adjusted training data sets were compared with those of the testing datasets. The results revealed that the mean average percentage error (MAPE) after adjusting the demand was 5.43% to 6.22% for the training and 9.55% to 10.33% for the testing dataset, which is lower than that of the traditional forecasting method when there was no intervention. In summary, the adjustment forecast could increase forecast accuracy by 42.39% and 39.42% for 20- and 40-feet containers, respectively.
Intervention; Qualitative forecasting method; Reefer containers; Third party logistics providers; Time series forecasting
The growth of container use in marine transportation has steadily increased in recent years due to economic development in Asia (Diaz et al., 2011). As the world’s largest archipelagic country, the distribution of agricultural and food products in Indonesia is normally conducted through sea transportation (BPS-Statistics Indonesia, 2018). According to the Indonesia Cold Chain Association, the number of logistics providers in Indonesia is still growing, especially third party logistics providers (3PLs), who provide shipping services with reefer containers (Gandi, 2016). The delivery of food products is carried out from Java to Eastern Indonesia using temperature control to maintain product quality. However, the accumulation and repositioning of empty containers have become major issues for container ports and could be worsened by the growing trade imbalances emerging among trading routes. Having too many emptied containers in certain locations is an indicator of other planning issues in the shipping business (Epstein et al., 2012), the first of which is an imbalance in container demand among different regions. The second problem is multiple sources of uncertainty, such as market conditions, and the third problem relates to the handling and sharing of operations information. These problems affect every third party logistics provider, increasing their operational costs as well as creating a shortage of containers and reducing their income (Pradita and Ongkunaruk, 2019). Previous research on container demand forecasting in the area of reefer containers has been as follows: Wang and Meng (2019) compared three forecasting models for the number of inbound of containers, while Tang et al. (2019) studied container throughput forecasting at Lianyungang Port and Shanghai Port by comparing the Gray model, triple exponential smoothing model, multiple linear regression model, and backpropagation neural network. Tangkham and Ongkunaruk (2019) suggested the container yard to establish a proper forecasting method for demand for containers in Thailand. In addition, Mo et al. (2018) studied container throughput forecasting using the group method of data handling (GMDH) neural network. They also studied a hybrid forecasting model (SARIMA), support vector regression (SVR), back-propagation (BP) neural network, and genetic programming (GP). There are two major forecasting methods. First, the qualitative forecasting method is analyzed based on the opinions, judgments, and past performance of forecasting experts (Arvan et al., 2019). The factors that influence demands for commercial products are sales promotions, the introduction of a new brand, store reform, and aggressive marketing (Meneghini et al., 2018). The type of qualitative forecasting using an expert adjustment or intervention is more appropriate when demand is highly sensitive to sales promotion (Min, 2008; Trapero et al., 2015; Fildes et al., 2019). Furthermore, online marketing demand prediction based on user/consumer reviews and promotional marketing using bigdata was investigated by Chong et al. (2017). The higher the forecast accuracy, the more the bullwhip effect is reduced (Jaipuria and Mahapatra, 2014). While qualitative forecasting is only used when the amount of historical data is limited, quantitative forecasting is more commonly used among practitioners. The types of most commonly used quantitative forecasting are time series, regression model (Taylor and Letham, 2018), Autoregressive Integrated Moving Average (ARIMA) (Min, 2008; Jaipuria and Mahapatra, 2014; Dhini, 2015), Seasonal Autoregressive Integrated Moving Average (SARIMA) (Farhan and Ong, 2018; Mo et al., 2018), Artificial Neural Network (ANN) (Jaipuria and Mahapatra, 2014; Dhini et al., 2015); and Multinomial Logit Model (MNL) (Lubis et al., 2019). In circumstances of promotion or irrational events, combining qualitative and quantitative methods could be implemented to increase forecast accuracy (Min, 2008; Jaipuria and Mahapatra, 2014; Khamphinit and Ongkunaruk, 2016; Chong et al., 2017). We hypothesized that the intervention forecasting approach would provide better forecasting accuracy than other time series alternatives, such as ARIMA and SRIMA. The primary advantage of the aforementioned techniques (i.e. BP, SVM, ANN) is a high mathematical precision; however, the disadvantages of calculation complexity and intensive data requirements deter practitioners from utilizing these techniques. In contrast, the time series technique requires less data to operate and is intuitively simple, yet it still provides an acceptable level of mathematical precision. Therefore, the time series technique was selected for use in this study. Moreover, the experience level of the forecasters will affect the results of their forecasting: the greater the forecaster's expertise, the greater the adjustment will correctly represent the real insights and improve forecast accuracy (Trapero et al., 2015; Arvan et al., 2019; Fildes et al., 2019). The combination of qualitative forecasting via the adjustment of sales data by experts as well as quantitative forecasting via time series using a software package could increase forecast accuracy by 46.14%, 22.53%, and 56.42% for three kinds of instant noodle products (Khamphinit and Ongkunaruk, 2016). Therefore, the intervention forecast method would perform well within the reefer circulation context, as in the present Indonesian 3PL case.
The findings showed that the MAPE values from the intervention forecasting approaches outperformed the traditional time series approach in this particular case. Specifically, the intervention forecasting approach revealed benefits such as having data adjusted by human insight twice. The first data adjustment acted as “a screening tool” to identify and correct all irregular demand points prior to feeding this “cleaned” data into the quantitative calculation. This step helped avoid the “garbage in/garbage out” phenomenon. The second intervention adjustment from experts acted as a “reality check tool”, which put human’s insights back into the result from the mathematical models. Again, this step was comparable to the seasonality index concept in time series forecasting, which adds a seasonal characteristic back to de-seasonalized forecast demand to represent the reality of each season. While the seasonal index might have static pattern the intervention by expert is more event-specific and more intuitive. Given that the reefer demand forecast is very sensitive to any weekly or even daily events, we can conclude that the combination of human insight from the qualitative method and mathematical precision from the quantitative method helped increase its forecasting accuracy. Overall, with more reliable forecasting, container procurement officers can then enhance both their circulation efficiency and customer service levels.
In regard to the impact of different product categories, the intervention forecasting method also yielded very similar results for both the 20- and 40-feet containers. For both container sizes, non-adjusted demand exhibited a higher forecast error than the adjusted demand for both sizes. As a result, the adjustment forecast helped increase the forecast accuracy by 42.39% and 39.42% for the 20- and 40-feet containers, respectively.
For implementation purposes, irregular events can be defined as repeatable or non-repeatable. Non-repeatable events can be disregarded as outliers, since they occur accidentally; however, repeatable events should be investigated further since they might happen again in a cyclical manner or should be included in associated forecasting models (e.g. regression). For example, natural disasters and subsidized projects from the government impact only impact forecast errors once, since such events tend not to occur in the future; however, irregular events, such as Ramadhan, in this case, will occur again. Thus, these social events should be systematically recognized and integrated into decisions to improve forecasting accuracy.
It is important to note that data limitations existed in this study, such as, for example, a lack of available quantitative data prior to 2016 as well as only having a single expert to provide insights on the qualitative side. In order to gain more power to apply these findings to other cases, repetitive research or longitudinal research is required. In addition, future research could be extended for comparison against other product categories, comparisons of different product life cycles, or comparisons with other forecasting methods, such as ANN or the regression method (Jaipuria and Mahapatra, 2014; Dhini et al., 2015).
The authors would like to thank the company for kindly providing information for this research. Also, we thank the anonymous reviewers who gave significant suggestions to help us improve our manuscript.
Arvan, M., Fahimnia, B., Reisi, M., Siemsen, E., 2019. Integrating Human Judgement into Quantitative Forecasting Methods: A Review. Omega, Volume 86, pp. 237–252
Box, G.E.P., Jenkins, G.M., Reinsel, G.C., Ljung, G.M., 2008. Time Series Analysis: Forecasting and Control. 4th Edition. Hoboken, New Jersey: John Wiley & Sons, Inc
BPS-Statistics Indonesia, 2018. Sea Transportation Statistics 2018. CV Dharma Putra, Jakarta, Indonesia
Chong, A.Y.L., Ch’ng, E., Liu, M.J., Li, B., 2017. Predicting Consumer Product Demands via Big Data: The Roles of Online Promotional Marketing and Online Reviews. International Journal of Production Research, Volume 55(17), pp. 5142–5156
Davydenko, A., Fildes, R., 2013. Measuring Forecasting Accuracy: The Case of Judgmental Adjustments to SKU-Level Demand Forecasts. International Journal of Forecasting, Volume 29(3), pp. 510–522
Dhini, A., Surjandari, I., Riefqi, M., Puspasari, M.A., 2015. Forecasting Analysis of Consumer Goods Demand using Neural Networks and ARIMA. International Journal of Technology, Volume 6(5), pp. 872–880
Diaz, R., Talley, W., Tulpule, M., 2011. Forecasting Empty Container Volumes. The Asian Journal of Shipping and Logistics, Volume 27(2), pp. 217–236
Epstein, R., Neely, A., Weintraub, A., Valenzuela, F., Hurtado, S., Gonzalez, G., Beiza, A., Naveas, M., Infante, F., Alarcon, F., Angulo, G, Berner, C, Catalan, J, Gonzalez, C, Yung, D., 2012. A Strategic Empty Container Logistics Optimization in a Major Shipping Company. International Journal of Interfaces, Volume 42(1), pp. 5–16
Farhan, J., Ong, G.P., 2018. Forecasting Seasonal Container Throughput at International Ports using SARIMA Models. Maritime Economics & Logistics, Volume 20(1), pp. 131–148
Fildes, R., Goodwin, P., Onkal, D., 2019. Use and Misuse of Information in Supply Chain Forecasting of Promotion Effects. International Journal of Forecasting, Volume 35(1), pp. 144–156
Gandi, I., 2016. Mitigating High Logistics Cost. In: UN ESCAP Regional Conference for Logistics Service Provider, Jakarta, Indonesia, pp. 49–55
Jaipuria, S., Mahapatra, S.S., 2014. An Improved Demand Forecasting Method to Reduce Bullwhip Effect in Supply Chains. Expert Systems with Applications, Volume 41(5), pp. 2395–2408
Khamphinit, T., Ongkunaruk, P., 2016. Combining Qualitative and Time Series Forecasting to Increase the Forecasting Accuracy for Instant Noodle Sales in Thailand. International Journal of KnE Life Science, Volume 3(3), pp. 1–5
Lubis, H.A., Pantas, V.B., Farda, M., 2019. Demand Forecast of Jakarta-Surabaya High Speed Rail based on Stated Preference Methods. International Journal of Technology, Volume 10(2), pp. 405–416
Meneghini, M., Anzanello, M., Kahmann, A., Tortorella, G.L., 2018. Quantitative Demand Forecasting Adjustment based on Qualitative Factors: Case Study at a Fast Food Restaurant. International Journal of Systems and Management, Volume 13(1), pp. 68–80
Min, J., 2008. Forecasting Japanese Tourism Demand in Taiwan using an Intervention Analysis. International Journal of Culture, Tourism, and Hospitality Research, Volume 2(3), pp. 197–216
Mo, L., Xie, L., Jiang, X., Teng, G., Xu, L., Xiao, J., 2018. GMDH-Based Hybrid Model for Container Throughput Forecasting: Selective Combination Forecasting in Nonlinear Subseries. Applied Soft Computing, Volume 62, pp. 478–490
Pradita, S.P., Ongkunaruk, P., 2019. Business Process Analysis and Improvement for a Third Party Logistics Provider in Indonesian Cold Chain Logistics. In: IOP Conference Series: Materials Science and Engineering, Volume 526
Talib, A., 2017. Tuna and Skipjack (A Management Review of Potential Resources in Indonesian Fisheries). Journal of Agribusiness and Fisheries, Volume 10(1), pp. 38–50
Tang, S., Xu, S., Gao, J., 2019. An Optimal Model based on Multifactors for Container Throughput Forecasting. KSCE Journal of Civil Engineering, Volume 23(9), pp. 4124–4131
Tangkham, K., Ongkunaruk, P., 2019. Business Process Analysis for a Container Depot Service Provider in Thailand, 2019. In: International Conference on Engineering, Science, and Industrial Applications (ICESI), Tokyo, Japan, pp. 1–5
Taylor, S.J., Letham, B., 2017. Forecasting at Scale. The American Statistician, Volume 72(1), pp. 37–45
Trapero, J.R., Kourentzes, N., Fildes, R., 2015. On the Identification of Sales Forecasting Models in the Presence of Promotions. Journal of Operational Research Society, Volume 66(2), pp. 299–307
Wang, Y., Meng, Q., 2019. Integrated Method for Forecasting Container Slot Booking in Intercontinental Liner Shipping Service. Flexible Services and Manufacturing Journal, Volume 31(3), pp. 653–674