Trigger-based retraining (based on performance degradation monitoring and data drift detection) can replace scheduled full retraining, directing computational resources precisely where they are truly needed