The diversity of characteristics among Micro, Small, and Medium Enterprises (MSMEs) often poses a challenge for effective policy formulation, where a one-size-fits-all approach fails to distinguish the specific needs of each business. This study aims to overcome the limitations of administrative segmentation by applying a new behavior-based framework, namely LSE (Longevity, Scale, Employment), to characterize the typology of MSMEs in the city of Semarang. The research method adapts the principles of Recency, Frequency, Monetary (RFM) into business metrics: Longevity (resilience), Scale (maturity), and Employment (socio-economic impact). Using data sourced from the Semarang City Cooperative and MSME Office, this study analyzed 8,804 valid MSME data points through a K-Means Clustering algorithm optimized with the Elbow Method. The results identified three distinct clusters: Sustainable Businesses, Accelerative Businesses, and Stagnant Micro Businesses. These findings validate the effectiveness of the LSE model in mapping business heterogeneity and recommend a paradigm shift in policy towards targeted and relevant interventions for each segment.