SUPPLY CHAIN DEMAND PREDICTION: THE EFFECT OF DEMAND INSTABILITY IN THE FRAMEWORK OF MARKETING

Demand sharing and demand forecasting are extremely beneficial for supply chain managers since they provide a good source of information that can be utilized for planning and decision making. When it comes to the supply chain, demand forecasting is the foundation for a great deal of administrative choices, including demand planning, order fulfilment, production planning, and inventory control. This study identified the model that is most suited to the various amounts of volatility, the impact of demand instability within the context of marketing by making use of supply chain demand forecast. Over the course of the years 2023 and 2024, the quantitative research was carried out in the Department of Management at North South University in Bangladesh. In order to collect data, we have taken into consideration the corporate sector. We make it a priority to collect as much information as we can in order to guarantee that the findings are reliable, accurate, and applicable to other businesses that are comparable. For the purpose of forecasting volatile demand, the research makes use of a hybrid regression time series model known as HR-ARIMA. The primary focus of the study is on the relationship between price and demand. When compared to the link between total demand and price, the authors discovered that there was a higher correlation between price and demand uplift due to promotions. Additionally, when it came to promotional uplifts, they built a piecewise regression model. There are considerable disparities between promotional and non-promotional times, with promotion considerably contributing to CoV, according to the findings of the study, which analyzes consumer demand data over a period of 108 weeks. Throughout the course of time, sales are classified into three distinct groups: low, moderate, and high. There are considerable disparities between promotional and non-promotional times, with promotion considerably contributing to CoV, according to the findings of the study, which analyzes consumer demand data over a period of 108 weeks. Throughout the course of time, sales are classified into three distinct groups: low, moderate, and high. Our research revealed that the exponential smoothing with covariate (ETSX) method has a poor performance when it comes to forecasting volatile demand series, however the ARIMA with covariate (ARIMAX) method is effective. Both support vector regression (SVR) and dynamic line-ear regression (DLR) models are capable of producing reliable forecasts across a wide range of demand categories, each of which has a unique value of the coefficient of variation (CoV).