Model Drift in Risk Score occurs when a scoring model starts performing differently from how it performed during development or validation. In MSME lending, borrower behaviour, GST patterns, bureau usage, industry stress, digital payments, and macro conditions can change over time. A variable that once predicted default may become weaker, or a new risk pattern may emerge. For example, a model trained during stable market conditions may understate risk during sector-wide stress. This matters because approvals, pricing, monitoring, and early-warning systems may depend on the score. The common caution is to monitor score distribution, bad rates, overrides, rejected cases, and variable stability regularly rather than assuming the model remains accurate forever.