Chapter 1 Resources
American Public University
COURSE SCMG305: Global Demand Mgt.
Introduction & Overview
Demand forecasting is a precise science and art rather than an intuitive process, as has been demonstrated in chapter 1 on Demystifying Demand Forecasting in the book Demand-driven forecasting: A structured approach to forecasting by Charles W. Chase (2013). Although the chapter elucidates the importance of data-driven approaches over personal judgment in improving the accuracy of the demand forecast, it creates interest in several areas that require further reading for enhanced understanding and broadening knowledge in demand forecasting. The chapter promotes the use of big data analytics and discourages personal intuition and simplified statistical approaches when predicting the demand in companies with enormous product or service portfolios. It notes that oversimplification of data analytics, ignoring irregular inputs like seasonality, random events, and extraneous factors delivered overambitious or subdued demand forecasts that did not help supply chains (Chase, 2013). In turn, two sources that augmented the knowledge gathered from this chapter were identified. Both are peer-reviewed journal articles that report on research studies. One deals with the intervention forecasting approach for improving demand forecasting accuracy while the other compares statistical approaches to demand forecasting with machine learning approaches, which are emerging due to technological advancements in analytics. In this respect, this discussion delves into the keywords used to retrieve the sources from the internet before summarizing each source. After that, a summary of what I have learned from these sources is highlighted.
Key Search Terms
The search terms used to identify and retrieve included models of demand forecasting and advanced technologies in demand forecasting. These key search terms were selected because the knowledge gained from Chapter one left gaps and raised curiosity in how much demand forecasting had advanced over time and the good practices that were being adopted by successful demand forecasting practitioners. Google Scholar was the search engine used to locate and retrieve relevant peer-reviewed research articles on the said topic. The search was limited to sources published in the last 5 years, thus yielding two articles, Pradita, Ongkunaruk, & Leingpibul (2020) and Spiliotis et al. (2020), both published in 2020. These articles were selected for their currency, relevance, and new knowledge.
The first source was authored by Sintia Putri Pradita, Pornthipa Ongkunaruk, Thaweephan (Duke) Leingpibul in the International Journal of Technology in 2020. It is a 10-page peer-reviewed journal article.
Summary of Content
This addresses the challenges in achieving accuracy in demand forecasting that continued to be experienced by companies with a global presence. It focuses on the challenges encountered in forecasting demand in reefer containers commonly used in the international trade of perishable products that require refrigeration during transportation and storage. The authors proclaim that reefer containers are in short supply due to ineffective forecasting practices leading to perpetual shortages. Using Indonesia as a setting for a field study aimed at identifying demand patterns and irregularities, Pradita, Ongkunaruk, and Leingpibul (2020) used data from several touchpoints in the container value chain, which incorporated allowances for extraneous variables, such as oil price increases and natural disasters to create demand prediction models. Precisely, they developed demand prediction models that combined qualitative and quantitative approaches, in which human insights were incorporated into mathematical approaches and compared them with traditional models. Their eventual model contained an intervention parameter that accounted for demand forecasting errors identified in the traditional models that were being employed by an Indonesian third-party logistics firm. In the end, their adjustment forecast improved the accuracy of the demand forecast for 20-feet and 40-feet reefer containers by 42.39% and 39.42%, respectively. They concluded that conducting an adjustment forecast that accommodates the mean average percentage error improved demand forecasting accuracy significantly. Such adjustments accounted for irregular events that influenced demand unpredictably. In turn, the article recognized the importance of a seasonality index in demand forecasting accuracy improvement and how this should be used to view the human insights that are determined qualitatively.
The second source was authored by Evangelos Spiliotisa, Spyros Makridakisb, Artemios-Anargyros Semenogloua, and Vassilios Assimakopoulos and published in Operational Research journal in 2020.
Summary of Content
This article focuses on the challenges of accuracy in demand forecasting in irregular demand scenarios that are erratic and intermittent. Although Spiliotisa et al. (2020) acknowledge that statistical methods of demand prediction are well established and have become common practice, they noted that machine learning approaches were increasingly being utilized to forecast demand in the daily stock-keeping units (SKUs). However, Spiliotisa et al. (2020) observed that using statistical methods to predict daily demand was challenging often delivering suboptimal results. In turn, their study compared the performance of statistical and machine learning demand forecasting methods to identify this was more accurate at low cross-sectional levels and contained less bias. After testing several statistical and machine language methods, the authors revealed that some machine learning methods, such as neural networks and regression trees delivered higher accuracy and lower bias compared to the well-established statistical methods like Croston’s method and Syntetos-Boylan Approximation. However, they noted that although the computation times for machine learning approaches were longer than those of statistical methods, they could be shorted significantly through cross-learning, making the computation times significantly shorter than those of statistical approaches. However, the authors advised on the need for further research in this area, particularly in the cross-learning across different machine language approaches and datasets. This is because the study focused on cross-learning using the series-by-series approach and therefore, did not reveal information about other cross-learning approaches. Besides, not all machine learning approaches achieved increased accuracy from cross-learning, which required further investigation to determine such differences.
The two sources added valuable information to what I had learned about demand forecasting from chapter 1 of the textbook. Firstly, I learned that human insights can contribute to the accuracy of demand forecasting when they are analyzed scientifically and presented as qualitative data rather than simply personal hunches. However, qualitative data needed to be collected from the perspectives of several participants in the demand forecasting process. Secondly, I learned that technological advancements were improving the accuracy of statistical methods, which are preferred as the objective demand forecasting approach because they eliminate bias and minimize demand prediction errors. In this regard, machine language approaches were helping improve the accuracy of demand forecasting by performing complex computations using large historical data sets from numerous demand-determination touchpoints.
In conclusion, demand forecasting is a fast-developing discipline inspired by the increased complexities in business structures and operations, unpredictability in market environments, and the ferocity of competition. Therefore, it should be objective by being data-driven rather than intuition-based. However, although statistical methods provide a scientific and evidence-based approach to demand forecasting, technological advancements were transforming the demand forecasting technicalities being employed in firms. Consequently, firms needed to predict demand accurately by incorporating all the variables that influenced the demand of their products and services, including those at the lowest granular level.
Chase, C. W. (2013). Demand-driven forecasting: a structured approach to forecasting. John Wiley & Sons.
Pradita, S. P., Ongkunaruk, P., & Leingpibul, T. D. (2020). Utilizing an intervention forecasting approach to improve reefer container demand forecasting accuracy: A case study in Indonesia. International Journal of Technology, 11(1), 144-154.
Spiliotis, E., Makridakis, S., Semenoglou, A. A., & Assimakopoulos, V. (2020). Comparison of statistical and machine learning methods for daily SKU demand forecasting. Operational Research, 1-25.