Prompt Performance Evaluation Rubric

# Prompt Performance Evaluation Rubric

You are a prompt quality assurance specialist. Design a complete evaluation rubric for comparing prompt versions for [TASK TYPE].

## Evaluation Dimensions

Define 6 evaluation dimensions with scoring scales:

**Accuracy (1-5):** Does the output contain correct information and meet the task requirements?

**Consistency (1-5):** Does the prompt produce similar quality output across 10 different inputs?

**Format compliance (1-5):** Does the output match the specified format on every run?

**Edge case handling (1-5):** Does the prompt handle unusual inputs without failing or hallucinating?

**Efficiency (1-5):** Is the output appropriately concise — no unnecessary padding or repetition?

**Robustness (1-5):** Does output quality hold when the model is updated or the input phrasing varies slightly?

## Scoring Method

Define the evaluation process:
- Minimum sample size for a reliable score: how many test inputs per dimension
- How to handle scoring disagreements when multiple evaluators review
- Threshold score that a prompt must reach before moving to production

## Comparison Template

Design a side-by-side scorecard for comparing Prompt A vs. Prompt B across all 6 dimensions, with a weighted total and a final recommendation.

## Regression Baseline

Explain how to establish a performance baseline for the current production prompt and how to detect when a new version is a statistically meaningful improvement vs. noise.