# Evaluation

## Protocol

- Data split:
- Selection data:
- Confirmation data:
- Random seed:
- Metrics:
- Threshold:

## Results

| Model | Primary metric | Secondary metric | Compared with baseline |
| --- | --- | --- | --- |
| Baseline |  |  |  |
| Candidate |  |  |  |

## Error analysis

- Largest errors:
- Groups or periods with weaker performance:
- Sensitivity to threshold or assumptions:

## Decision

- Recommendation:
- Conditions under which the result should not be used:
- Remaining uncertainty:

## Done when

- The test set was not used for tuning.
- Error patterns are discussed, not just the best score.
- The recommendation follows from the stated decision metric.