A retail company runs a daily sales forecasting model that has been producing inaccurate predictions for the last 3 weeks. Each day, an AWS Glue job merges model inputs, the model's predictions, and actual daily sales into files that are stored in Amazon S3. The ML team identifies that the inaccuracies are caused by shifts in the feature value distributions and needs a solution that will detect such data drift in the future with the least operational overhead. Which approach meets this requirement?

Amazon SageMaker Model Monitor is designed specifically for detecting data and model quality issues, including data drift. By creating a data quality baseline, it learns the expected feature distributions. Enabling emitmetrics allows Model Monitor to send drift metrics to Amazon CloudWatch, where an alarm can be configured to notify the team when significant drift occurs, providing an automated and low-overhead solution. "Model quality baseline" is incorrect because model quality focuses on performance metrics like accuracy, not feature distribution shifts. Amazon SageMaker Debugger is for debugging training jobs, not for monitoring deployed model predictions or data drift. Using Amazon CloudWatch to manually inspect logs for data drift is reactive and lacks the automated detection capabilities of Model Monitor, leading to high operational overhead.