Your QA team scores calls every week. Agents see a number, sometimes a comment, and almost never a change in what they do next. The operation keeps missing the same targets, because the form measures whether an agent said the greeting rather than whether the customer’s problem got solved. Call center quality assurance usually breaks at three specific points: the scorecard measures the wrong things, two scorers grade the same call differently, and findings never reach a coaching conversation. This post covers how to fix all three, plus who should own the function. Every change here is one you can start this quarter.
Design a call center quality assurance scorecard around outcomes
Most scorecards get built backward. Someone lists everything an agent is supposed to say, assigns points, and the total becomes the score. An agent can hit 95 percent on that form and still leave the customer calling back tomorrow.
Start from the KPIs the operation is accountable for, then work out which agent behaviors move them. On the statewide education benefits program Mpathic designed, built, and operates, the contractual targets are 80 percent of calls answered within 30 seconds, abandonment at 8 percent or lower, average handle time of 10 minutes or less, and first call resolution of 80 percent or higher. Mpathic helped define those KPIs rather than inheriting them, which is the part worth copying.
An outcome-weighted form asks different questions. Did the agent verify the right record. Did they use the knowledge base article that states the rule. Did they resolve inside their own authority instead of transferring. Did the case note leave the next agent able to pick the contact up cold. Script adherence stays only where a phrase is contractually required, weighted as a checkbox, not a grade.
ISO 18295-1:2017 specifies service requirements for customer contact centres, including performance metric specifications. Read it as a coverage checklist, then keep your form short enough that reviewers complete it honestly. The constraint is reviewer time, not the field count your software allows.
Calibrate your scorers before you defend a single score
A score agents distrust is usually a score that moves with the scorer. When two reviewers grade the same call and land 12 points apart, the number is describing the reviewer, and agents will treat their results as a lottery.
Calibration is a standing meeting with a fixed procedure. Take one recorded contact a week. Every scorer grades it independently, then you compare line by line and spend the time only where you disagree. Where the disagreement traces to an ambiguous criterion rather than to judgment, rewrite the criterion. Evaluation tooling supports that sequence: in Amazon Connect, a reviewer opens a custom evaluation form against a contact, reviews the recording or transcript, and can save a draft before submitting. See Amazon Connect agent performance evaluation forms for configuration details.
Front-load the cadence. In a lead qualification and appointment setting engagement, Mpathic committed to daily calibration for the first 30 days after go-live and a weekly QA review through the end of month three. New programs generate almost all of the ambiguity. Steady-state programs need maintenance.
Publish the calibration spread. Agents accept a scoring system faster once they see the scorers being checked too.
Route every finding into coaching within the week
A QA finding has a shelf life. Scored on Monday and delivered in a monthly review, it arrives as criticism of a call nobody remembers.
Recording has to cover every contact you intend to review, and that is a configuration decision made up front. In Amazon Connect, a “Set recording and analytics behavior” block has to sit in the contact flow before the contact reaches an agent, and reviewers need security profile permissions to open what was captured. See Amazon Connect recording setup for the configuration steps. On the state education benefits program, the reported KPI set names call recording and transcription for management review, QA, coaching, and up-skilling as one deliverable, which is the right way to write it down: the recording exists in order to change what an agent does next. Mpathic then built Amazon QuickSight dashboards on that program for real-time KPI visibility including performance by individual representative, reviewed with the agency daily and weekly.
Put a service level on your own QA function. A finding gets a coaching conversation within five business days, the coach writes down what was agreed, and the next review checks that behavior before scoring anything new. If a criterion never produces a conversation, take it off the form. You are paying reviewer hours to produce a number nobody uses.
Decide who owns quality assurance inside the operation
QA fails quietly when it belongs to everybody. Supervisors score when the queue allows, so the weeks that most need review are the weeks nobody reviews.
The staffing pattern on that statewide program is a dedicated role: eight full-time customer success representatives, one full-time trainer and QA specialist, and one customer success manager. One person owns training and quality assurance for the whole team, embedded in the operation rather than reporting from a separate department. Mpathic ran a train-the-trainer engagement with the program administrator, then took over ongoing training and quality assurance. Treat that ratio as a reference point rather than a rule. If the volume does not support a full-time reviewer, buy a fraction of a dedicated one, through an internal shared role or staff augmentation, instead of spreading the duty across supervisors.
That engagement bills as a monthly recurring charge rather than per hour, which puts the incentive on staff quality, retention, and continuous improvement instead of billable hours. If your outsourced customer support contract pays for hours, QA is a cost your provider absorbs. If it pays for outcomes, QA is how your provider protects its margin. Published results on that program include customer satisfaction of 4.88 out of 5.0 on after-call surveys and first call resolution of 93 percent against a contractual 80 percent. The build and operating detail sits in the state agency contact center case study.
Start with one change
Pull your current scorecard and count how many criteria measure whether the customer’s problem got solved. If it is fewer than half, rebuild the form against your KPIs before you score another call. Then put one calibration session on the calendar, have every scorer grade the same contact independently, and publish the spread. Call center quality assurance earns agent trust through consistency and follow-through, not through better wording on the form.
If you are rebuilding a QA function, or want it run by a team that staffs it as a real role, talk to our team. We will start with your scorecard and your KPIs.
Frequently asked questions
What is quality assurance in a call center?+
Call center quality assurance is the practice of reviewing recorded or live customer contacts against defined criteria, then using what the review finds to change agent behavior. A working program has three parts: a scorecard tied to the operation’s KPIs, calibrated scorers who grade the same contact the same way, and a route from each finding into a coaching conversation.
Why is quality assurance important in a call center?+
Because it is the only mechanism that connects a specific customer interaction to a specific improvement. Reviews show whether agents resolve inside their own authority, use the knowledge base correctly, and leave usable case notes, all of which drive first call resolution and CSAT. Without QA, an operation can watch a metric slip for months without knowing which behavior caused it.
What should a call center quality assurance form include?+
Criteria drawn from the KPIs the operation is accountable for, weighted toward resolution rather than script recitation: record verification, correct use of the knowledge base, resolution within the agent’s authority, and case notes a colleague can pick up cold. ISO 18295-1:2017 includes performance metric specifications useful as a coverage checklist.
How often should call center QA calibration happen?+
Front-load it. In one Mpathic engagement design, calibration ran daily for the first 30 days after go-live, then a weekly QA review continued through the end of month three. New programs generate most of the scoring ambiguity, so heavy early calibration settles the form while steady-state operations need maintenance rather than intensity.
Who should own quality assurance in an outsourced contact center?+
A named role inside the operation, not a duty shared among supervisors. On a statewide education benefits program, Mpathic staffs one full-time trainer and QA specialist alongside eight customer success representatives and a customer success manager, with that person owning ongoing training and quality assurance for the team.
