Metric Selection Worksheet
Use this before reporting model quality.
Decision context
What action will the model support?
Who uses the result?
Who is affected?
Target
Target column:
Positive class, if classification:
What does a positive prediction mean?
Error costs
| Error | Plain-language meaning | Possible harm or cost |
|---|---|---|
| False positive | ||
| False negative |
Candidate metrics
| Metric | Keep or reject? | Reason |
|---|---|---|
| Accuracy | ||
| Precision | ||
| Recall | ||
| F1 | ||
| ROC-AUC | ||
| PR-AUC | ||
| MAE | ||
| RMSE |
Chosen metric
Primary metric:
Why this metric matches the decision:
Secondary metric or check:
Threshold note
If this is classification, how might the threshold change the action?
Limitation note
What should a reader not overclaim from this metric?