Which approach is used to detect when a deployed model's performance degrades due to changes in input data distribution?

Correct answer: Data drift monitoring

Option D is correct because data drift monitoring is the specific MLOps technique for continuously tracking the statistical properties of a model's input features in production and alerting when the distribution of incoming data diverges significantly from the training distribution, which signals that model performance may degrade and retraining may be required. Option A, A/B testing, is used to compare the performance of two or more model versions against each other using live traffic, not to detect distributional shifts in input data. Option B, hyperparameter tuning, is a training-time optimization process for improving model accuracy by searching over configuration parameters, and it has no role in monitoring a deployed model's input distribution. Option C, cross-validation, is a training-time technique for estimating how well a model generalizes to unseen data by partitioning training data into folds, and it is not applicable to monitoring data distribution changes in a production deployment.