What triggers model retraining?

• A. When there is more data available.
• B. When the model architecture changes.
• C. When the training time decreases.
• D. Performance degradation or drift detection.

Answer: (D)

Retraining is triggered when performance falls short or when the data environment changes. In production, you monitor the model’s metrics (like accuracy, AUC, precision/recall) and watch for drift in the input data or in the relationship between features and the target. When degradation or drift is detected, retraining with fresh data helps restore accuracy and keep the model aligned with current conditions. Simply having more data available isn’t by itself a trigger, and architectural changes or a shorter training time aren’t automatic reasons to retrain—those relate to design adjustments or efficiency, not ongoing performance signals.