Support QA Scorecard template.

What's inside — the exact file

This is the complete template, not a sample. The rows are worked examples — replace them with your data (and delete them before any platform import).

What this template is for

Most support QA fails the same way: two reviewers score the same ticket differently, agents call it subjective, and the program dies. This rubric fixes the calibration problem — six weighted criteria, each with written definitions for scores of 1, 3, and 5, so two reviewers reading the same conversation land on the same number.

The six criteria (accuracy, completeness, tone, process compliance, ownership, and writing quality) with their weights are a proven starting point — adjust the weights to your values, but never remove the written score definitions. The definitions are the calibration; the numbers are just arithmetic.

How to use it, step by step

1. Adapt the definitions to your context. Rewrite each score definition with your product's specifics — what 'accurate' means for a billing dispute vs. a bug report. Keep the 1/3/5 anchor structure.
2. Weight by what you actually value. The weights should sum to 100 and reflect your priorities — accuracy usually deserves the most. If tone matters more than process in your brand, say so in the numbers.
3. Calibrate before you score anyone. Two reviewers score the same five tickets independently, then compare. Where scores diverge by 2+, the definition is ambiguous — rewrite it. Repeat until divergence is rare.
4. Sample fairly and consistently. Score 3–5 tickets per agent per week, randomly sampled with maybe one flagged ticket. Cherry-picked reviews destroy trust faster than no reviews.
5. Review with the agent, not at them. The rubric turns feedback from opinion into shared criteria. Agents should be able to self-score with the same sheet and predict their result — that's when QA starts improving quality instead of measuring it.
6. Track scores over time and against CSAT. Rubric scores that don't correlate with customer satisfaction over a quarter mean the rubric measures the wrong things — revisit the criteria.

What each column means

Common mistakes to avoid

• Scoring without written definitions — the moment scores are contestable, the program is politics.
• Ten criteria instead of six: reviews take too long, so they stop happening. Fewer, weightier criteria survive contact with real workloads.
• Using QA scores punitively from day one — score anonymously for the first month while calibrating, or you'll never get honest buy-in.
• Never scoring AI agent conversations — automated resolutions deserve the same rubric; that's how you find where the AI needs better content or tighter guardrails.

Questions, answered

What's a good QA score target?

After calibration, healthy teams average around 4.0–4.5 out of 5 with the definitions written honestly. A team averaging 4.9 has soft definitions, not perfect quality — tighten the score-5 anchors until excellence is distinguishable.

How is this different from CSAT?

CSAT measures how the customer felt; QA measures whether the work was right. They disagree constantly — a wrong answer delivered charmingly scores high CSAT, and a correct 'no' scores low. You need both lenses, and this rubric is the internal one.

Should this rubric score AI conversations too?

Yes, unchanged. Sample your AI agent's automated resolutions weekly with the same six criteria. It's the fastest way to find the gap between 'the bot answered' and 'the bot answered well' — and it makes the human-vs-AI quality comparison honest.

This is a Market Disrupt worksheet, not a vendor file. Import-format templates follow each platform's documented layout as of July 2026 — platforms evolve, so validate against current documentation before a large import.