Forecasting Preliminary Order Cost to Increase Order Management Performance: A Case Study in the Apparel Industry

Introduction

The changing dynamics of today's business world point out that data analyzing should be taken into account with importance in all processes of companies. The use of companies' existing data for future forecasts opens a significant horizon for company executives, both in terms of decision-making processes and strategy setting (Guimaraes and Paranjape, 2021).

In the labor-intensive ready-to-wear apparel industry, order costs have a complex structure consisting of many elements. The formation of unpredictable costs on orders makes detailed cost estimation studies both costly and time consuming. The use of data mining techniques in determining the preliminary cost of order and achieving successful results will provide the company with a great competitive advantage and increase order management performance (Pujitha and Venkatesh, 2020).

In some cases, in order to be cautious, the order is missed due to cost estimated higher than its normal value. . In addition, loss occurs as a result of the actual cost value above the estimated cost due to unforeseen extra costs. Both situations cause a decrease in performance for manufacturers producing for global brands in the market.

Due to the fact that the factors that make up the costs consist of many variables and the unpredictable interaction between the variables, difficulties are encountered in estimating the costs, hence the forecasts show low accuracy (Adhikari, Bisi, and Avittathur, 2020). In addition, the dynamic functioning of the sector does not allow high precision calculations in terms of both cost and time constraints. For these reasons, developing high-performance models for order cost estimation is of great importance for supply chain efficiency and company performance (Ngai,et al, 2014).

In general, the entire supply and production process in the textile industry is labor intensive. For this reason, labor costs consisting of human factors have an intense effect on the cost. The fact that the production process is heavily influenced by the human factor rather than the mechanical steps, causes the formation of actual cost data to deviate greatly from the arithmetic and linear preliminary cost estimates (Pujitha and Venkatesh, 2020). In addition, the involvement of external suppliers that are difficult to control over causes deviations from the estimates that are made by linear methods. Based on this approach, in this study, the cost estimation system used by an apparel manufacturer company with arithmetic methods is compared with the cost estimates that will be created by using machine learning and artificial neural networks for the same orders. To the best of our knowledge, this is the first study on forecasting preliminary -order costs in the apparel industry and fills an important gap in the literature.

The article content includes a review of literature of related subject in Section 2. Methodology of the study is explained in Section 3. Structure of the model, performance of the forecasting models and evaluation of results are explained with a case study in Section 4. Conclusion of the study and further research directions are given in Section 5.