Agent performance: measure what agents say, not just how fast they say it

What to measure per agent

Two axes: throughput (calls, AHT) and quality (QA score, adherence, sentiment). Measuring throughput alone rewards fast rudeness; measuring quality alone rewards perfectionists who take forever. You need both.

Building a scorecard that's fair

Fair scorecards have three properties: weighted criteria that reflect what actually matters, calibrated scoring that removes reviewer bias, and transparent reasoning so agents see why they lost points. Auto-QA delivers all three by default; manual QA needs deliberate work to.

Coaching workflow: from score to 1:1

1. Data: per-agent gaps identified.
2. Recording: pull calls that show the gap.
3. Plan: 1–2 focus areas per agent per month, with specific behaviors.
4. Follow-up: measure the trend, not one call.

Best-call libraries & onboarding

Real examples of your playbook done right — new hires learn from actual calls, not slide decks. Best-call library is the highest-leverage output of auto-QA in most operations.

Auto-QA: how 100% coverage changes the team

The QA team shifts from listening (~10 people on samples) to analysis (~1–2 analysts calibrating criteria and coaching). QA cost drops to ~20% of manual review. Work becomes more valuable, not less.

Avoiding Goodhart's law

"When a metric becomes a target, it ceases to be a good metric." Don't tie bonuses to a single scorecard number — measure the trend across multiple axes, and validate with sampled listening.