Call Center Quality Assurance: The Framework, and How to Close the Gaps It Finds
By Dave Wilson · 7 min read · 13 July 2026
Call center quality assurance is the process of listening to, or reading transcripts of, agent calls, scoring them against a fixed rubric, and using that score to catch problems and coach agents before they turn into churn. Done well, it is the earliest warning system a contact center has. Done as a compliance exercise, a monthly form filled in and filed away, it produces a spreadsheet of scores and no measurable change in how calls actually go.
What call center quality assurance actually measures
A QA program scores a sample of calls against a rubric that usually covers three layers: compliance (did the agent follow required disclosures and verification steps), process (did they follow the correct workflow for the issue), and communication (tone, empathy, active listening, ownership). Most programs weight these three unevenly, and the weighting says a lot about what the business actually values.
The best QA rubrics score five to ten specific, observable behaviours, not vague categories like "professionalism." "Did the agent state a specific next step and timeframe before ending the call" is scoreable. "Was the agent professional" is not, because two QA reviewers will score the same call differently.
Building a QA scorecard: what to include
A scorecard that actually produces consistent scores across reviewers covers five categories, each with an example of what "yes" looks like on a real call.
| Category | What it checks | Example criterion |
|---|---|---|
| Compliance | Required disclosures and verification were completed | Confirmed the customer's identity before discussing account details |
| Process adherence | Correct workflow for the issue type was followed | Logged the case in the correct queue with the required fields |
| Communication | Tone, clarity, and active listening | Acknowledged the customer's specific concern before offering a solution |
| Resolution | Issue was actually resolved, or escalated correctly | Confirmed the customer understood the next step before ending the call |
| Ownership | Agent took responsibility rather than deflecting | Explained a policy limit without blaming "the system" or hiding behind process |
Practise the scenario your QA score flagged, try it now, no sign-up needed.
Practise the scenario your QA score flaggedHow many calls should you QA per agent
There is no single right number, but the pattern that holds up across contact centres is roughly two to four calls per agent per month for a stable, low-risk queue, moving to weekly review for new hires in their first ninety days or agents on a coaching plan. Scoring every interaction, which many QA platforms now automate, catches more, but only helps if someone actually acts on what it finds. A perfectly scored spreadsheet nobody reviews is not better than a smaller sample somebody actually coaches from.
Why QA programs stall: the scoring stops, the coaching does not start
The failure mode is consistent across most contact centres that run a QA program and still see the same mistakes every quarter. Calls get scored. Scores get logged. Nobody closes the loop by turning a specific, recurring gap into a specific practice session before the next batch of calls is reviewed.
A QA scorecard that shows the same few agents scoring low on de-escalation, month after month, is not a scoring problem. It is a training gap the QA program correctly found and nobody fixed, because reading a low score does not teach the behaviour the score was measuring.
Practise the scenario your QA score flagged, try it now, no sign-up needed.
Practise the scenario your QA score flaggedFrom QA finding to closed gap: turning scores into practice
The contact centres that actually move their QA scores over time treat a low score as the start of a training loop, not the end of a reporting cycle: pull the specific calls that scored low on a given category, name the exact moment the call went wrong, then have the agent rehearse that exact moment, an angry customer refusing to calm down, a policy limit that needs saying without sounding dismissive, until the response holds up under pressure.
That rehearsal step is where most programs quietly give up, because getting a supervisor to role-play a convincing angry customer for every agent who needs it does not scale. AI roleplay for customer service fills exactly that gap: agents practise the specific scenario their QA score flagged against a realistic AI customer, with a transcript and score after every run, so the next QA cycle is scoring a rep who has already rehearsed the thing that failed last time. Pair it with customer service de-escalation scripts for the language behind the most commonly flagged category, or call center training for how to structure practice beyond QA-flagged gaps.
Frequently asked questions
What does QA do in a call center?
A QA analyst listens to or reads a sample of agent calls, scores them against a fixed rubric covering compliance, process, and communication, and reports the results back to team leads so they can coach specific gaps. In smaller teams, a supervisor usually does this directly rather than a dedicated QA analyst.
What skills do you need for call center quality assurance?
Strong analytical skills to score consistently against a rubric rather than a gut feeling, clear written and verbal communication to explain a score in a way an agent can actually act on, and enough front-line experience to know what a good call sounds like. The strongest QA analysts have usually worked the phones themselves.
How many calls should you QA per agent?
Roughly two to four per month for a stable, low-risk queue, rising to weekly for new hires in their first ninety days or agents on an active coaching plan. Automated scoring of every call catches more, but the number only matters if someone reviews the findings and turns them into coaching.
What's a good call quality score?
Most contact centres treat 85 to 90% as a healthy baseline on a well-built rubric, but the trend matters more than the number. An 82% average that is rising because a specific coaching gap closed is a better sign than a static 90% nobody is checking the components of.
What's the difference between QA and coaching in a call center?
QA measures what happened on past calls against a fixed rubric. Coaching is what a team lead does with that measurement, working with an agent to close a specific gap before the next batch of calls is scored. A QA program without a coaching loop attached to it is just a report; the score only creates value once someone turns it into a specific practice session.
A QA scorecard is only as useful as what happens after the score is logged. The contact centres that improve are not the ones with the most sophisticated rubric or the highest sampling rate. They are the ones that treat every low score as an assignment: here is the exact moment that went wrong, now go rehearse it before it comes up on a real call again.
See it in Oliver
Ready when you are
Practise before the next real conversation.
No sign-up required. Pick the scenario, add your name, and speak with a realistic AI partner. Get a scorecard and transcript after every call.