Your question is Regression Metric for Asymmetric Cost. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What evaluation metrics would you use for a regression model where overpredicting is significantly more costly than underpredicting?
Explain how you would measure asymmetric error, which standard metrics you would retain or avoid, and how you would select an operating threshold or decision rule. Discuss how you would validate that the metric reflects the actual cost of prediction errors.