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?