A Guide to Quality Assurance (QA) in Customer Support

Metrics like first response time and resolution time tell you how fast your support team works. They do not tell you how well. A ticket resolved in ten minutes means nothing if the answer was wrong, the tone was dismissive, or the customer left more confused than when they started. That gap—between how quickly your team responds and how well they actually help—is where quality assurance lives.

Quality assurance (QA) in customer support is the practice of systematically reviewing interactions to ensure they meet your standards for accuracy, empathy, professionalism, and effectiveness. It is the mechanism that turns good intentions into consistent execution. Without it, you have hope. With it, you have a system.

Why QA matters

Consistency across the team

Without QA, each agent develops their own style. Some are thorough, some are brief. Some are warm, some are clinical. Some follow policy precisely, some improvise. Your customers should get a roughly similar experience regardless of which agent handles their ticket. QA is how you make that happen instead of leaving it to chance.

Identifying training needs

QA reviews surface specific skill gaps that speed-based metrics never will. An agent with fast resolution times but low empathy scores needs coaching in one area. An agent with poor accuracy needs coaching in a different one. Without QA, those gaps stay hidden until they show up in customer complaints—which is too late.

Protecting the customer experience

Every interaction your team has shapes your customer's view of your business. QA ensures that experience is reliably good, not dependent on which agent happens to answer. That consistency builds the trust that drives retention.

Continuous improvement

QA data reveals systemic patterns. If multiple agents struggle with the same question, your knowledge base needs updating. If tone scores are dropping across the team, your communication guidelines need refreshing. QA turns individual reviews into organization-wide insights—the feedback loop that lets you improve continuously.

Building your QA programme

Choose your criteria

Your scorecard defines what "good" actually looks like. Select five to eight criteria covering the most important aspects of a support interaction. Common ones:

Accuracy — Was the information correct? Did the agent follow the right process?

Completeness — Did they fully address the customer's question, or leave loose ends?

Empathy — Did they acknowledge the customer's situation and show understanding?

Tone — Was the communication style appropriate for the channel and situation?

Efficiency — Did they resolve the issue without unnecessary back-and-forth?

Policy compliance — Did they follow your procedures and company policies?

Grammar and formatting — Was the response well-written and easy to read?

Not every criterion matters equally. Accuracy might count for 30 percent. Empathy for 20 percent. Grammar for 5 percent. Weight them based on what actually drives customer satisfaction in your business.

Define your scoring scale

Keep it simple. A three-point scale works well for most teams:

• Meets expectations (2) — The agent handled this criterion well
• Needs improvement (1) — There is a specific area where the agent could do better
• Does not meet expectations (0) — The agent missed this significantly

Some teams use five points for more granularity. Three is usually sufficient and reduces ambiguity for reviewers (five-point scales sometimes become a form of decision paralysis where everything ends up a 3).

For each criterion and score level, document specific examples. What does "meets expectations" look like for empathy? What does "needs improvement" look like for accuracy? Concrete examples ensure different reviewers apply the scorecard consistently.

Who conducts reviews

In small teams, team leads or managers run QA as part of their regular responsibilities. In larger teams, dedicated QA roles provide the most consistent review. Some teams use peer review—agents reviewing each other's work—which has the advantage of being educational for the reviewer and reducing the "surveillance" feeling of top-down QA. Self-review (agents reviewing their own interactions first) also builds self-awareness.

The best approach usually combines these: peer review for learning, manager review for accountability, self-review for development.

Sampling strategy

Reviewing every interaction is not practical. You need a sampling approach that gives you real coverage without overwhelming your reviewers.

Random sampling provides an unbiased view of typical performance. A common target is five to ten reviews per agent per month. Targeted sampling supplements this: review tickets with low CSAT scores, escalations, complex multi-touch interactions, and interactions from new agents (higher sampling during onboarding). You can use ticket prioritisation criteria to identify high-value conversations worth reviewing more frequently.

As a rough guide: small teams (under 10 agents) aim for five reviews per agent per month; medium teams (10–30 agents) aim for four per month; large teams aim for three per month, supplemented by AI-assisted review where possible.

The review process

Set a regular cadence for reviews—weekly or biweekly. Consistency matters more than volume. Five reviews conducted reliably every week are worth more than twenty done sporadically.

Share QA results with agents within a week of the interaction being reviewed. Feedback on a three-month-old ticket has almost no impact. Feedback on a conversation from last Tuesday will actually change behavior.

Make feedback specific and actionable. "Be more empathetic" is not useful. "When the customer mentioned they'd been waiting a week, acknowledging that wait before jumping to the solution would have made the response feel more caring" gives them something concrete to work on.

Separate the coaching conversation from the scoring. QA scores should inform coaching, but the coaching itself should be supportive and forward-looking. Focus on what the agent can do differently next time, not on penalising past performance.

If multiple people conduct reviews, hold regular calibration sessions where everyone reviews the same interaction and compares scores. Discuss disagreements and align on how to apply the scorecard. Without calibration, different reviewers interpret criteria differently, and your data becomes unreliable. Use real-time support dashboards to track which reviewers are aligned and which need recalibration.

Using QA data

QA scores reveal each agent's strengths and development areas. Use this to create personalised coaching plans. Someone who scores well on accuracy but struggles with tone needs different coaching than someone who is warm but occasionally inaccurate.

Aggregate QA data across the team to identify patterns. If everyone scores low on completeness, the issue is likely systemic—the knowledge base is incomplete, or the team needs training on a product area. Compare QA scores against CSAT, first contact resolution, and escalation rates. Agents with high QA scores should generally have high CSAT scores. If they don't, either your QA criteria don't align with what customers actually value, or external factors (product issues, long wait times) are overshadowing agent quality.

Track QA scores over time. Are they improving, declining, or flat? Improvements validate your coaching efforts. Declines signal emerging issues. Flat scores might mean your coaching isn't landing or your targets need raising. Build custom reports and dashboards to visualize these trends so you can spot patterns quickly.

Common mistakes to avoid

Treating QA as punishment. If agents fear QA reviews, they'll be defensive rather than open to learning. Frame it as a development tool, not a disciplinary mechanism.

Reviewing only bad interactions. If QA only examines complaints and escalations, the data is negatively biased. Random sampling provides a balanced view of actual performance.

Inconsistent scoring. Without regular calibration, QA data is unreliable. Two reviewers scoring the same interaction differently undermines the whole programme.

Not acting on the data. QA without follow-up action wastes everyone's time. Every review should lead to either positive reinforcement or a specific improvement action.

Over-complicating the scorecard. Fifteen criteria with ten scoring options each creates analysis paralysis. Focus on what matters most.

Frequently Asked Questions

How long does a typical QA review take?

Most interactions average 3–7 minutes to review, depending on interaction length and scorecard complexity. A 5-minute email takes 5 minutes to evaluate. A 20-minute chat might take 8–10 minutes to score fairly.

What's the difference between QA and performance management?

QA is about quality and development. Performance management is about accountability. QA reviews identify skill gaps and coaching opportunities. Performance reviews determine compensation and advancement. You can use QA data to inform performance reviews, but QA itself should be non-punitive.

Can I use AI to automate QA reviews?

Partially. AI can flag interactions that fall outside normal parameters (unusual length, multiple escalations, repeated emotional keywords) and score some objective criteria like compliance checklist items. But subjective criteria—empathy, tone, completeness—still require human judgment. Most teams use AI to identify which interactions to review, then conduct manual review on those. It saves time and ensures human eyes on what matters most.

How do I handle QA for different channels?

Different channels need different criteria. A phone call emphasizes tone and verbal clarity. An email emphasizes completeness and grammar. A chat emphasizes speed and brevity. You can use the same core scorecard but adjust weights and examples for each channel. Channels like social media naturally have a more conversational tone, so you might score formality differently there.

What should I do if an agent consistently scores low?

Have a coaching conversation first. Identify the specific skill gap (accuracy? empathy? process compliance?). Provide targeted training, additional resources, or mentoring. Set clear improvement expectations with a timeline. If scores don't improve after reasonable effort, you have a valid performance management discussion. But most agents improve with specific, supportive coaching.

How do I avoid QA being seen as surveillance?

Transparency helps. Tell your team how many interactions are reviewed per month, how the scorecard works, and how feedback is used. Involve agents in building the scorecard if possible. Frame it as "here's how we maintain quality" not "here's how we catch mistakes." And actually use the data to improve the team's experience—provide better tools, fix knowledge base gaps, adjust policies that are creating problems.

Should I review my own QA reviewer?

Yes. If managers conduct reviews, calibrate with other managers regularly. If you have a dedicated QA person, review a sample of their reviews to ensure consistency. The reviewer's consistency matters as much as the agents' quality.

How often should I update the scorecard?

Review it quarterly. Are agents consistently maxing out or bottoming out on certain criteria? Are there emerging issues the scorecard doesn't capture? Are certain criteria not matching what customers actually care about? Use that data to refine. But don't overhaul it constantly—stability matters for trend analysis.

Getting started

1. Define five to eight QA criteria based on what drives customer satisfaction in your business
2. Build a simple scorecard with clear examples for each criterion and score level
3. Train reviewers and calibrate on sample interactions
4. Start with random sampling—five reviews per agent per month
5. Share feedback within one week of the interaction being reviewed
6. Review aggregate data monthly to identify team-level trends
7. Iterate the scorecard based on what you learn

QA is not a project with an end date. It is an ongoing practice that improves your team continuously. Start simple, be consistent, and let the data guide your evolution. If you're using Relentify Helpdesk, you get built-in QA workflows with customizable scorecards and analytics that track quality trends over time—removing the spreadsheet work so you can focus on actual coaching.