Quality and Call QA: People
Objective. Open People and compare agent-level quality only after aligning call mix, complexity, sample size and reviewer calibration - never read a bare agent score as a verdict.
People rolls scores up to the agent and team. It answers 'who needs support and who is strong?' - but only once the comparison is fair, which is the whole skill of reading this screen.
See it in the demo
What this screen answers
The People screen aggregates evaluations by agent and team. It answers 'where should coaching attention go?' A CSM uses it to prioritise support and spot agents trending down; Sales uses it to show quality can be seen at the person level - carefully positioned, because this is also the screen where unfair comparison does the most damage to trust with a customer's frontline.
The task
Scope to the team and period, then before comparing anyone, align the basis: are agents being judged on similar call mix, complexity and channel, on enough reviewed interactions, and by calibrated reviewers. The task is to produce a fair read - flag who genuinely needs support and who is strong - and to mark any agent whose score rests on too few or uncalibrated reviews as 'insufficient evidence', not as low performing.
How to read it
Compare two agents only after aligning work type, complexity, period, tenure, sample size and reviewer calibration. The classic misread is ranking agents by raw score as if the number were a clean performance verdict. It is not - a low score on a handful of hard calls scored by one uncalibrated reviewer is noise dressed as a rating. A small sample per agent makes individual scores especially volatile. Read People as a prompt to investigate, not a league table, and always check how many reviewed interactions sit behind each agent before drawing a conclusion about them.
Honest limits
Agent scores inherit sampling bias and calibration gaps and can be unfair at low volumes. This screen does not prove an agent's true quality, only how their sampled, reviewed interactions scored. Using it for consequential people decisions needs the customer's own process and sign-off. Confirm with product how people-level roll-ups and thresholds are configured and entitled; the demo shows seeded data.
Interpretation check
An agent's score is only comparable after aligning call mix, complexity, sample size and reviewer calibration - a low score on a few hard, uncalibrated reviews is noise, not a performance verdict.
In the field
KnackGlobal (RCM): comparing two coders' quality is only fair after aligning claim type and complexity - one handling the hardest payers will score differently for reasons that are not performance.
Practical task
Open the People screen, pick two agents with different scores, and write a fair-comparison note: what you would need to align before ranking them, and which one you would actually investigate first and why - flag either as insufficient evidence if the sample is thin.
Self-check
Agent A scores below Agent B. Is A the weaker performer?
Not without aligning call mix, complexity, period, tenure, sample size and reviewer calibration. Raw score order is not a fair ranking.
Why are low-volume agent scores risky to read?
Because a handful of reviewed interactions makes the score volatile - one hard call or one uncalibrated reviewer swings it. Mark it as insufficient evidence, not low performance.
What should you check behind each agent before concluding anything?
How many reviewed interactions sit behind the score, their mix and complexity, and whether the reviewers were calibrated.
Can this screen drive a consequential people decision on its own?
No. It prompts investigation and coaching priority; consequential decisions need the customer's own process and sign-off, not a demo roll-up.
Verified vs demo. Screen is live in the demo at /#/people; exact controls and figures are SME-validation-pending.