Call Quality Management (QA) Basics: Scorecards and Coaching
How to run call quality assurance: choose what to measure, build a scorecard, sample calls fairly, calibrate reviewers and coach agents.
Every contact centre believes it provides good service, but few can show how they know. Quality management, often called QA, is the practice of listening to a sample of conversations, scoring them against agreed criteria and using the results to improve. Done well, it makes training specific and fair. Done badly, it becomes an exercise in ticking boxes that agents resent.

This guide covers what to measure, how to design the scorecard, how to sample, how to keep reviewers consistent and how to coach. It sits alongside Contact Center KPIs Explained and Measuring CSAT After Calls.
Decide what quality means for you
Before writing a form, agree what a good call is in your business. Different teams value different things: a clinic values calm and accuracy; a sales team values clear next steps; a support line values resolution on the first contact. Ask three questions.
- What do customers need from this call?
- What must never go wrong, such as a mis-stated policy or a missed identity check?
- What behaviours help most in getting it right?
Typical quality dimensions
- Opening and identificationGreeting, name, and checking who is on the line.
- Understanding the needListening, clarifying questions, not interrupting.
- AccuracyCorrect information, correct process, correct record afterwards.
- Resolution and next stepsThe caller leaves knowing what happens next.
- Tone and mannerCourtesy, pace and empathy appropriate to the situation.
- ClosingSummary, thanks and a polite end.
Build the scorecard
A scorecard is a list of items that a reviewer marks pass, fail or not applicable. Keep it short enough to complete in a few minutes and specific enough that two reviewers would mark the same way.
Designing a scorecard
- List behavioursStart from the dimensions above and write each as something you can hear.
- Remove opinionReplace "was friendly" with "greeted the caller by name" or "used a polite closing".
- Weight the itemsGive more weight to items that affect outcomes or risk.
- Mark critical itemsChoose the few failures that make the whole call a fail, such as giving wrong information on a legal matter.
- Set a pass markDecide the score that counts as acceptable.
- Test itScore ten calls with two reviewers and rewrite items that cause disagreement.
Fewer, clearer items
Long forms tire reviewers and bury what matters. Between eight and fifteen items usually suffices. Every item should lead to an action: if no one would change training because of it, drop it.
Critical items
A critical item is one where failure makes the call unacceptable whatever else happened, so it zeroes the score. Use them sparingly. If most items are critical, none is.
Weak vs strong scorecard items
| Weak item | Stronger item | |
|---|---|---|
| Courtesy | Agent was nice | Agent used the caller's name and a polite closing |
| Understanding | Agent understood the need | Agent restated the request before acting |
| Accuracy | Gave good information | Information matched the current policy page |
| Wrap-up | Updated the record | Recorded the outcome and next step in the call record |
Sample fairly
You cannot review every call by hand, so choose carefully.
- Random sample. A fixed number per agent per period, picked without bias, gives a fair picture of normal work.
- Targeted sample. Add calls that are very long, ended in a complaint, show low customer feedback or contain risk words. These reveal problems.
- New staff. Review more calls from people in their first weeks.
Tell agents how calls are chosen. Secret sampling feels like surveillance; transparent sampling feels like support. Be clear that recorded calls are used for quality and training, and give callers notice of recording as required where you operate. See Call Recording Notices.
Keep reviewers consistent
If one reviewer marks every call high and another marks every call low, scores mean nothing. Calibration fixes that: a few times a month, all reviewers score the same call independently, then compare and discuss differences. Where scores differ widely, the item wording is usually the problem, so rewrite it.
Track the spread between reviewers. A wide spread means they read the scorecard differently, and the team should agree a common interpretation before the scores are used for anything important.
Turn scores into coaching
Scores have no value until they change behaviour. Good coaching is regular, specific and based on real calls.
- Play a short section of a call with the agent, not just the score.
- Ask the agent to say what they would do differently.
- Agree one or two actions, not ten.
- Check them in the next review.
- Praise what worked, using examples.
Avoid public rankings and avoid using QA as the only basis for discipline. Agents who fear the form will script themselves into stiff calls. Use scores to find patterns across the team as well: if many agents miss the same item, the training, the script or the tools are at fault, not the individuals. Practice calls and role-play help; see Practice Calls: Training Agents Before They Go Live.
A sustainable weekly rhythm
Quality programmes die when they depend on heroic effort. A small team can keep one going with a simple weekly pattern: reviewers score a fixed number of calls early in the week, a short team huddle shares one pattern and one good example, and each agent gets a ten-minute one-to-one with a single action to try. Once a month, hold a calibration session and look at the trend for the whole team.
Keep a short log of changes you made because of QA, such as a revised script line, a new knowledge-base article or an extra step in a process. When agents see that review findings lead to real fixes, they stop treating QA as a trap and start bringing their own examples.
Balance QA with customer outcomes
A call can pass every item and still leave the customer unhappy, and the reverse is also possible. Combine QA scores with other signals: customer satisfaction replies, first-contact resolution, repeat calls and complaints. Listen to the calls behind the unhappy answers first, because a specific caller with a recording attached is easier to act on than an average. See First Call Resolution and What Contact Center Supervisors Should Monitor.
Doing it in Call
Call keeps reviewing and coaching on one screen:
- Review queue. Calls waiting for review are listed, and a reviewer scores each item as pass, fail or not applicable. A failed critical item sets the score to zero. The score shows as you go, and a comment explains what to do differently.
- Automatic scoring. Items that can be expressed as words said or not said can be scored automatically against rules you set. Keep rules narrow; a blank item is better than a guessed one.
- Coaching sessions and agent scores. See each agent's results and where the team is weakest, and book coaching sessions.
- Calibration. A spread between reviewers shows when scorecards are read differently.
- Customer feedback. Satisfaction and net promoter results, with the calls behind unhappy answers attached, and a way to create an after-call survey.
The setup steps are in Building Quality Scorecards in Call. Supervisors can also hear live calls; see Live Listen and Coaching for Supervisors in Call.
Frequently asked questions
How many calls should I review per agent?
Enough to see a pattern, not so many that coaching never happens. Many teams review a few calls per agent per period and add more for new staff or when a problem appears. Agree a number you can keep up.
Should QA scores affect pay?
Be cautious. Tying pay tightly to scores encourages box-ticking and arguments over marks. Many teams use scores mainly for coaching and use them in pay decisions only with calibrated, transparent criteria.
Can software score calls automatically?
Partly. Items that can be expressed as words said or not said can be checked automatically. Tone, empathy and judgement still need a person, so automatic scoring supports human review rather than replacing it.
What is calibration?
A regular session where reviewers score the same call independently and then compare. It exposes unclear items and makes scores comparable between reviewers.
How do I avoid agents feeling watched?
Explain the purpose, show how calls are chosen, give feedback promptly and focus on development. Include positive examples, and give agents a way to challenge a score.