<![CDATA[Business environments in the last five years have been shaped by accelerating change and external uncertainty, prompting organizations and researchers to re-evaluate forecasting methodologies used to support managerial planning. This literature review aims to map the evolution of business forecasting methods during 2020–2025 by synthesizing empirical evidence from primary international journals indexed in Scopus and national journals accredited within Indonesia’s research indexing ecosystem SINTA Index. The findings show that many organizations, especially small-to-medium entities, still rely on linear statistical methods, such as moving averages, exponential smoothing, and ordinary least squares regression, due to their low data requirements and rapid implementation through spreadsheets. However, these methods often generate higher forecast errors when data characteristics shift over time, fluctuate sharply, or contain structural disruptions. In contrast, international studies indexed in Scopus Q1–Q2 journals report increasing adoption of non-linear machine learning models and deep learning temporal predictors, including LSTM and GRU, which demonstrate stronger ability to learn complex patterns and maintain predictive resilience under rapid demand changes. Hybrid forecasting models emerge as the most prominent research trend, offering a balanced integration between linear statistical baselines and non-linear learning refinements to enhance forecast robustness, stability, and contextual risk adaptation. The review highlights a substantial implementation and research gap in Indonesia, particularly in large-scale business settings, signaling future opportunities for real-time adaptive and AI–statistical hybrid evaluations using local enterprise datasets.]]>
