An ML model is deployed in production. The model has performed well and has met its metric thresholds for months. An ML engineer who is monitoring the model observes a sudden degradation. The…

Question

Options

• ALack of training data
• BDrift in production data distribution
• CCompute resource constraints
• DModel overfitting

• A
  5% (1)
• B
  90% (19)
• D
  5% (1)

Explanation

When a model performs well for months and then suddenly degrades, the most common root cause is data drift-the statistical distribution of incoming production data has shifted away from the distribution the model was trained on. Real-world data changes over time (seasonal patterns, economic shifts, new customer behaviors), and a model trained on old data becomes less accurate. Option A (lack of training data) would have been a problem at model creation time, not months later. Option C (compute resource constraints) would cause latency or timeouts, not a drop in prediction accuracy below metric thresholds. Option D (model overfitting) is a training-time problem that would be detected during evaluation before deployment, not appear suddenly after months of good production performance.

No community discussion yet for this question.