# Ticket QA Grader

> Grade support agent responses against your quality rubric automatically. Score tickets on tone, accuracy, resolution speed, and policy compliance — then surface coaching opportunities without reading every ticket yourself.

Source: https://cotera.co/solutions/support/ticket-qa-grader

---

- **Team:** Support
- **Tools:** Zendesk, Google Sheets
- **Difficulty:** medium
- **Setup time:** 10 min
- **Time saved:** 3 hrs/week

## How it works
1. **Pulls resolved tickets** — Fetches recent tickets from Zendesk with full conversation threads, agent names, and resolution metadata.
2. **Scores each dimension** — Evaluates every response against your rubric — tone, accuracy, resolution speed, and policy compliance.
3. **Flags coaching opportunities** — Identifies specific tickets where agents missed the mark, with concrete examples and suggested improvements.
4. **Builds a scorecard** — Exports per-agent averages, team trends, and top performer highlights to Google Sheets.

## Example requests
- Grade the last 50 resolved tickets from this week using our standard rubric
- Show me which agents have the lowest tone scores and pull example tickets
- Find tickets where our refund policy was applied incorrectly

## Prompt

```markdown
## Task

Grade support ticket responses against a quality rubric. For each ticket, evaluate the agent's response on tone, accuracy, resolution speed, policy compliance, and customer effort. Produce a scorecard with specific feedback and coaching suggestions.

## Input

The user provides:
1. A set of resolved tickets to grade (from Zendesk, Intercom, or a CSV)
2. Their quality rubric or scoring criteria (or use the default rubric below)
3. Optionally, a date range or agent name to filter by

**Example:** "Grade the last 50 resolved tickets from this week. Use our standard rubric: tone (1-5), accuracy (1-5), first-response time, and policy compliance (pass/fail)."

## Context

### Default Quality Rubric

**Tone (1-5):**
- 5: Warm, empathetic, personalized. Customer feels heard.
- 3: Professional but generic. No errors but no personality.
- 1: Cold, dismissive, or robotic. Copy-paste without context.

**Accuracy (1-5):**
- 5: Correct solution on first try. No follow-up needed.
- 3: Mostly correct but required clarification or a second message.
- 1: Wrong answer, misleading information, or missed the actual question.

**Resolution Speed:**
- Fast: Resolved in 1 response
- Normal: Resolved in 2-3 responses
- Slow: 4+ responses or escalation needed

**Policy Compliance (Pass/Fail):**
- Followed refund/return policy correctly
- Used approved messaging for sensitive topics
- Properly escalated when required

### Grading Process

1. Pull resolved tickets from the specified source
2. For each ticket, read the full conversation thread
3. Score each dimension of the rubric
4. Flag any policy violations or coaching opportunities
5. Calculate per-agent averages and team averages
6. Identify top performers and agents needing coaching

## Output

**QA Summary**

**Tickets Graded:** [count]
**Average Score:** [X/5]
**Policy Compliance Rate:** [X%]

---

**Per-Agent Breakdown:**

| Agent | Tickets | Avg Tone | Avg Accuracy | Avg Speed | Compliance | Overall |
|-------|---------|----------|-------------|-----------|------------|---------|
| [Name] | [count] | [score] | [score] | [speed] | [pass/fail rate] | [avg] |

---

**Coaching Opportunities:**

- [Agent]: [Specific feedback with ticket example]
- [Agent]: [Specific feedback with ticket example]

---

**Top Performers:**

- [Agent]: Consistently high scores on [dimension]. Example: Ticket #[ID]
```

## Related reading
- [Best AI Customer Support Tools](https://cotera.co/articles/best-ai-customer-support-tools.md)
- [AI Customer Support for Shopify](https://cotera.co/articles/ai-customer-support-shopify.md)