Support Quality Assurance (QA) Framework

QA scorecard design determines what behaviors the support team optimizes for — so the scorecard criteria must reflect the actual customer experience and business outcomes you want to drive, not the behaviors that are easiest to measure. Scorecard structure: Critical criteria (immediate zero or case review — any fail here has policy or legal implications): accuracy of information given (never score a factually incorrect resolution as passing), privacy compliance (no PII shared inappropriately), and completion commitment (if the agent committed to a follow-up action, it was completed). Core quality criteria (scored 1–5 on each): Problem understanding (did the agent correctly identify the core issue, not just the surface symptom?); Solution quality (was the solution given the most effective available, fully addressing the issue?); Communication clarity (was the response clearly written, free of jargon inappropriate for the customer's technical level, and easy to follow?); Empathy and tone (did the agent acknowledge the customer's experience appropriately and maintain a warm, professional tone throughout?); Efficiency (was the agent appropriately efficient, avoiding unnecessary back-and-forth while not rushing at the expense of quality?). Avoid measuring mechanical behaviors that don't predict customer experience (greeting format, sign-off phrase, specific word count) — these create rigid compliance without genuine quality.

QA calibration is the process of aligning multiple QA raters on a shared interpretation of the scorecard criteria — ensuring that Agent A receiving an 87% QA score from Reviewer 1 means the same quality level as an 87% from Reviewer 2. Without calibration, QA scores are rater-dependent rather than performance-dependent — an agent's score fluctuates based on who reviewed them, not what they did. Calibration session format: monthly, 60-minute session with all QA reviewers and the Support Ops lead. The team selects 3–5 representative interactions ahead of the session (one clearly above-standard, two typical, one borderline). Each reviewer scores independently before the session. During the session, scores and rationale are shared and discussed. Where reviewers disagreed by more than 10 points on a criterion, the group discusses the scorecard language and agrees on the correct interpretation for that scenario — updating the QA playbook with a documented example. Calibration target: after calibration, inter-rater reliability (the average scoring agreement across all criteria between any two raters) should be > 85%. Below 75% indicates a fundamental disagreement about criteria definitions that the scorecard language must resolve.

QA programs become counterproductive when they are primarily used as performance management tools — agents optimize for passing QA reviews rather than genuinely helping customers, and managers use QA scores as the primary evidence in performance improvement plans. A constructive QA-to-coaching integration: QA as learning data, not judgment: frame QA scores as "quality data that tells us where to invest coaching effort" not as "grades that determine your worth." The QA reviewer and the agent review the same scored interaction together — the agent sees the same context the reviewer saw, making the score a shared reference rather than an external judgment. Strengths-based coaching: start every coaching session with what the agent did well, and why it worked (not as false flattery — as genuine identification of the specific decision or behavior that created a good outcome). Agents who understand their strengths can apply them more deliberately. Development-focused improvement: when a criterion score is low, the coaching question is "what would help you handle this more effectively?" not "why did you do it wrong?" — opening a problem-solving frame rather than a defensive one. QA trend tracking for team health: quarterly, review the distribution of QA scores across the team. An improving distribution means the QA and coaching system is working. A stagnant or declining distribution means the coaching approach itself needs redesign.