What metric is monitored in production?

• A. Prediction accuracy and input distributions.
• B. Model packaging size only.
• C. Training loss during development.
• D. Number of features in the dataset.

Answer: (A)

In production, the focus is on how well the model performs on real, live data and whether the data it sees matches what it was trained on. Monitoring prediction accuracy on recent predictions shows whether the model is still making correct decisions, while watching input distributions reveals if the data characteristics have shifted (data drift) and could degrade performance. Together, these metrics tell you when to retrain or adjust data pipelines.

The other options don’t reflect ongoing production behavior: training loss is a development metric used during model-building, not for live monitoring; model packaging size concerns deployment footprint rather than performance or data quality; the number of features is a fixed characteristic of the dataset and doesn’t indicate how well the model is performing in production.