To QA support tickets, define what “good” means on a scorecard, review interactions against it (ideally all of them, not a sample), calibrate scores across reviewers, and turn the results into coaching and fixes, closing the loop so quality actually improves.
To QA support tickets, define what “good” means on a scorecard, review interactions against it (ideally all of them, not a sample), calibrate scores across reviewers, and turn the results into coaching and fixes, closing the loop so quality actually improves. The goal is not to grade agents for its own sake. It is to make support quality measurable and then act on what it shows.
1. Define your quality standard
Before you score anything, agree on what a good interaction looks like. Translate that into concrete, inspectable criteria (accuracy, resolution, tone, compliance, and process) rather than a vague sense of “helpfulness.” Every criterion needs an owner who will act on it. A standard nobody uses is overhead.
2. Build a scorecard
Turn the standard into a weighted support QA scorecard: each criterion, its weight, and what counts as a pass, a fail, or an automatic fail. Reserve auto-fails for serious compliance or accuracy breaches. Keep the scorecard short enough that scoring is fast and consistent, because a bloated scorecard gets abandoned within a quarter.
3. Decide coverage: sample vs full
Traditional QA reviews only about 1 to 2 percent of tickets by hand, enough to coach a few agents but far too little to trust a trend or catch a rare compliance failure.1 If you can, move to AI auto-scoring that covers up to 100 percent of tickets, the way platforms like unitQ do, so your QA signal is complete and unbiased.1 If you are still manual, at least sample randomly and stratify by channel and agent so the sample is representative rather than convenient.
4. Score and calibrate
Score interactions against the scorecard, then calibrate: have multiple reviewers score the same tickets and reconcile the differences until scoring is consistent. Calibration is the step teams skip and the reason QA scores lose credibility, because an inconsistent score is worse than no score. With automated QA, calibration shifts to tuning the model against your rubric and spot-checking its judgments.
5. Coach on specifics, not averages
Use the scores to coach individual interactions, not monthly averages. “Here is the ticket, here is what the scorecard flagged, here is what good looks like” changes behavior. “Your QA score dropped two points” does not. Tie every coaching conversation to the exact criterion that failed.
See how unitQ compares on your data
A short demo, run on your own feedback.
6. Close the loop to the product
The highest-value tickets in your QA data are the ones where the agent did everything right and the customer was still unhappy, because the problem is the product, not the agent. Route those patterns to the teams that own the fix. This is where support QA stops being an HR exercise and becomes an early-warning system for the business.
7. Measure whether quality is improving
Track the scores over time, but watch the outcomes that matter more: first-contact resolution, CSAT, repeat-contact rate, and compliance exceptions. If QA scores rise while those outcomes hold flat, your scorecard is measuring the wrong things, so revisit step one.
FAQ
Score ticket QA at full coverage
Want ticket QA scored at full coverage and tied to product root cause? See how unitQ Support fits your QA program.
Related guides
Sources 1 references
SQM Group, "Automated vs. Manual QA: How to Improve Accuracy, Insights, and Cost Efficiency." sqmgroup.com. Accessed October 2026.