Forecasting plays a critical role in organisational decision-making, especially within supply chains for production planning, procurement, and inventory management. Inaccurate forecasts, which can deviate by 20–50%, contribute to global supply chain waste estimated at USD 1.1 trillion annually. The automotive industry is particularly vulnerable due to its long lead times, capital intensity, and complex supplier networks. Studies reveal that one-third of forecasts contain errors up to 30%, with some deviating by as much as 90%, often leading to inefficiencies, excess inventory, or shortages. Toyota’s downward production revision for 2024 illustrates how disruptions can quickly invalidate forecasts.
In Singapore, the rapid rise of electric vehicles (EVs) and shifts in brand performance, such as BYD surpassing Toyota in early 2025, highlight the limitations of models relying solely on historical data. These changes reinforce the need for agile forecasting systems.
This study considers forecast accuracy evident in the automobile industry and investigates the root cause of these inaccuracies. Then, it evaluates both traditional forecasting approaches (moving averages, exponential smoothing, and regression) to enhance forecast accuracy for some large automobile manufacturers, using publicly available data. We measure the accuracy based on forecast errors, mean absolute deviation (MAD), mean absolute percentage error (MAPE), and tracking signal (TS). Finally, we consider a simulation-based forecasting method to see whether higher forecasting accuracy can be achieved.