Support QA Worth Doing
Most QA programmes produce scores nobody acts on. The useful version reviews a small sample properly and changes something every week.
What to review
Not everything, and not randomly either. A blend works best.
A rubric people can agree on
Five criteria, scored simply, and every agent has read them before being scored against them.
Anything more granular gets debated rather than acted on.
Make it about the next reply
QA that produces a monthly score changes nothing. QA that produces one specific piece of feedback per agent per week changes the next reply.
Keep the score for trends and spend the conversation on a single concrete improvement. And publish the good examples: agents learn more from reading an excellent reply to a hard ticket than from any rubric.
What QA will surface that you did not expect
Run QA for a month and one pattern shows up in nearly every team: pairs of tickets where two agents answered the same customer about the same problem, sometimes with different answers.
It scores as a quality failure and it is not one. Neither agent could see the other ticket, and no amount of coaching fixes that. It is a queue problem wearing a quality costume.
Frequently asked questions
How many tickets should we review per agent?+
Five to ten a month for a random sample, plus every reopen and every low CSAT with a comment.
Should QA scores affect pay?+
Be careful. Scores tied to compensation get gamed, and the first casualty is honest self-reporting.
Who should do QA?+
A lead or a dedicated quality role. Peer review works for calibration and struggles when feedback gets difficult.
The quality failure nobody can coach
Two agents answering one customer differently is caused by a duplicate ticket, not by either agent.
Start free trial14-day free trial. No credit card required.