Call Quality Assurance: How to Review Every Call Instead of a Sample

In short: call quality assurance (QA) is a systematic review of agents' calls against a fixed scorecard, to give feedback and improve results. Most call centers review a small sample by hand. Automatic QA reviews every call against the same scorecard, and leaves the manager the calls that need attention.

Why a manual sample is not enough

A manager listening to one call spends at least the length of the call, plus a few minutes filling in the form. So in most call centers each agent is reviewed on a handful of calls a month. The problems:

• The sample is not representative: one great or terrible call decides the monthly score.
• Different reviewers, different scores: two managers give the same call different scores, and agents feel it.
• No picture of the team: you cannot tell which criterion is weak for everyone, not just for one agent.

How to build a good QA scorecard

A good scorecard is short, clear and tied to what brings results for the business.

1. Start from the outcome: what makes a call good for you? A close, a booked meeting, first-call resolution?
2. Choose 5 to 10 criteria: for example opening and introduction, needs discovery, presenting value, handling objections, summary and next step, courtesy and language.
3. Give each criterion a weight: in sales, needs discovery is worth more than a perfect opening.
4. Define what "pass" means: one sentence per criterion, with an example. "Asked at least two questions about the customer's need before presenting a price."
5. Add compliance items if needed: statements that must be said, for example in insurance and finance.

Example of a short scorecard for a sales call

Criterion	Weight	What is checked
Needs discovery	25%	Open questions about the situation and the need before presenting the solution
Presenting value	20%	Connecting the need that came up to the product
Handling objections	20%	Listening, clarifying the objection and a relevant answer
Close and next step	20%	A clear ask for commitment or a set follow-up time
Opening and language	15%	Introduction, courtesy, pace and clear language

Calibration: making sure scores are fair

Automatic QA needs calibration too. The process is simple:

1. Pick 20 to 30 varied calls.
2. Two managers score them by hand, without seeing the automatic score.
3. Compare: where is the gap large? Usually it is a criterion whose definition is unclear.
4. Sharpen the definition and run it again.

In TalkWiz a manager can give any call a manual score. The manual score is kept next to the automatic one without overwriting it, and every finding comes with a quote from the call, so it is easy to see why a score was given. Agents can mark a finding they think is wrong, and managers see how many findings were marked correct.

From sampling to reviewing every call

When every call is reviewed, the way you work with the data changes too:

• A review queue: instead of picking calls at random, the manager gets the low-score calls, angry customers, and calls where the status the agent marked does not match what happened.
• A team heat map: agents against criteria. You see right away whether a problem belongs to one agent or the whole team, and whether a group training is needed.
• Daily feedback for the agent: the agent sees their calls, what went well and what to improve, without waiting for the monthly feedback session.
• Alerts on urgent calls: a legal threat, churn risk or an angry customer reach the manager right away.

QA in a service center

In a service center the criteria are different: first-contact resolution, empathy, accurate information, managing expectations about timelines. The outcome is different too: "completed" or "unresolved" instead of "interested". The same principle works: a short scorecard, weights, calibration and a review of every call.

Next step

If you want to see how your scorecard looks on real calls, open a free account: the first 100 calls are analyzed on us. For a broader look at the field, read the complete guide to AI call analysis.