In production monitoring, why are input distributions tracked alongside accuracy?

• A. To monitor only the model's size.
• B. To monitor only training loss.
• C. To detect changes in input distributions that can affect performance.
• D. To eliminate data drift automatically.

Answer: (C)

Tracking input distributions in production is about detecting data drift. When the inputs that the model sees in the real world change over time, the relationship the model learned during training can no longer hold, even if accuracy on historic data remains high. Monitoring these distributions alongside accuracy gives an early warning system: you can spot shifts, trigger retraining or feature adjustments, and maintain reliable performance. It’s not about model size or training loss, and drift isn’t automatically eliminated—detection guides the appropriate response.