29/10/2026
Customer service QA
Customer service QA (quality assurance) is a support team's routine of reviewing a sample of conversations against a written standard, then fixing what falls short, such as a wrong fact, an off tone, or a problem left unsolved.
How does customer service QA work?
A manual QA review scores a sample of conversations, so it fits into a week. Each week, pull a random handful of conversations plus every one an AI agent couldn't answer, and score the two groups separately, so the unanswered ones don't hide inside the team's average.
Reviewers score each conversation against a short scorecard. A failed line leads to one of two fixes, coaching the person who replied or fixing the help center page the answer came from.
How do you calculate a customer service QA score?
Divide the points a conversation earned by the points possible, then multiply by 100, so each review becomes a number the team can track week to week.
QA score = points earned ÷ points possible × 100
- Points earned: the score each scorecard line received.
- Points possible: the maximum for every line that applies to the conversation.
- Left out: lines that don't apply, such as a refund line on a login question.
- Auto-fail: a critical line, such as a wrong fact, that fails the whole review.
Zendesk's QA documentation states the auto-fail rule plainly, "If a category is marked as Critical, a rating below 50% automatically fails the review, resulting in an overall score of 0%." Average the review scores for the week.
For example, a refund reply that gets the amount right, reads clearly, and stays polite, but skips the 30-day refund window and leaves the customer asking again, earns 6 of 10 points, or 60%.
Why does QA matter once an AI agent answers customers?
An AI agent repeats whatever its pages say, so one wrong page can become the same wrong answer for everyone who asks. QA finds that answer while it's still one page to fix.
Does a happy CSAT rating mean the answer was right?
No. A polite reply can earn a happy rating and still quote last year's policy.
CSAT is the customer's verdict on one conversation, while QA is the team's own check against its standard, including the conversations nobody rated.
The failure Wonderchat's support team sees most often is retrieval pulling an archived policy and answering from it correctly.
A reviewer catches that failure by opening the page the answer came from, and the fix then belongs on the page, not only in the reply. Wonderchat answers from a team's own help center and shows the source on every reply.
Frequently asked questions
What is quality assurance in customer service?
Quality assurance in customer service is reviewing support conversations against a standard the team writes down, such as accuracy, tone, and whether the problem was solved. Reviews lead to coaching for people and to fixes on the help center pages that answers come from.
What should a customer service QA scorecard include?
Keep it to four to six lines a reviewer can score in a few minutes, such as correct, clear, on tone, followed policy, and solved. Mark correctness as critical, so a wrong fact fails the review whatever the other lines score.
How do you QA answers from an AI agent?
Review a weekly sample plus every question the agent couldn't answer. Check each answer against the page it came from, and when that page is wrong or missing, fix the page, so the next answer comes out right.
