# Support QA rubric (template)

Edit the criteria, weights and examples, then upload this file to the GPT's Knowledge. Weights should add up to 100%.

## Criteria
### 1. Understanding (15%)
- 4: Restated the issue accurately and asked only the questions needed.
- 2: Partly understood; asked for information the customer already gave.
- 0: Answered a different question.

### 2. Accuracy (25%)
- 4: Every statement matches policy and knowledge base.
- 2: Minor inaccuracy that did not affect the outcome.
- 0: Wrong information that affected the outcome.

### 3. Completeness (15%)
### 4. Tone and empathy (15%)
### 5. Clarity (15%)
### 6. Ownership (10%)
### 7. Process (5%)
(Add 0/2/4 anchors for each, using real examples from your team.)

## Auto-fail rules
- (list the conditions that make a conversation fail regardless of score)

## Channel notes
- Chat: (for example, short messages are fine; greeting not required on every message)
- Email: (for example, greeting and sign-off required)
- Phone transcripts: (for example, do not score grammar)