MACHINE LEARNING EVALUATION WORKSHEET Fictional tool-library teaching example; no model was trained. Task / prediction moment: Owner / permitted use: Features available at prediction time: Label definition and delayed or missing outcomes: Approved data and permissions: Duplicates / related groups / chronology: Training / validation / held-out test roles: Fitted preparation / model / settings / versions: Baseline and acceptance criteria: Actual outputs and uncertainty: Fictional scorecard: TP=6, FP=2, FN=3, TN=9; total=20. Accuracy=(6+9)/20=75%. Precision=6/8=75%. Recall=6/9~=66.7%. Always-on-time baseline: 11/20=55% accuracy; late-class recall=0. These invented counts teach arithmetic, not real model quality. Error review / staff workload: Source and policy changes / monitoring / fallback: Human decision and unresolved questions: No reminder, fee, or borrowing restriction is authorized by this worksheet.