Context

VoltSim builds a regression model to predict electric scooter battery degradation so fleet operators can schedule maintenance. The issue is that only 58% of scooters have complete lab-test discharge curves, while the rest have partial field telemetry with missing temperature and load measurements.

Current Performance

The Problem

Leadership wants to use the model for maintenance planning across the full fleet, but performance degrades sharply when validation relies on incomplete data. You need to determine whether the current validation approach is trustworthy and how to validate the model when experimental or simulation data is missing or only partially observed.

Requirements

1. Interpret what the metric gap between complete and incomplete data suggests.
2. Identify likely sources of validation bias caused by missing experimental or simulation inputs.
3. Propose a validation strategy that can still estimate real-world model reliability.
4. Recommend model or data improvements to reduce risk before broader deployment.
5. Explain what additional checks you would run before approving use in production.

Constraints

• Maintenance decisions above a 25-cycle error can trigger unnecessary battery replacement.
• Collecting new lab data takes 6 weeks and costs $900 per battery pack.
• Fleet operations need a go/no-go decision within 2 weeks.