Manual QA samples a few calls per agent per month and the sample is rarely representative. Xperia scores every call that reaches it, against a scorecard you define rather than a fixed template.
Calls are evaluated across Call Opening, Personalisation, Communication, Listening, Tone, Probing, Responsibility, Hold Procedure, Upselling and Closing. Within those sit the individual checks — self-introduction, delayed opening, customer name usage, empathy, effective listening, clarity, grammar, accountability, avoiding jargon, reducing fillers, professionalism, pace, effective probing, open-ended questions, proactive problem solving, adaptability, verification and follow-up, and closing greetings, among others.
Each category carries a point weight and the total must come to 100, validated as you edit. A sales floor can weight Upselling and Probing heavily while a support desk weights Listening and Closure. Individual parameters also carry a partial-credit floor, so a missed check costs a defined fraction of its points rather than always zeroing.
Scoring models are scoped per department, so one account can run entirely different scorecards for its support queue, its collections team and its outbound sales floor — each with its own parameters and weightings, reported separately.
Describe a new quality signal in plain English and Xperia generates the evaluation prompt and a human-readable definition for it, then quotes the additional per-call cost before you commit. Deleting a parameter checks every department's scoring model first and warns you which weightings would break.
Beyond scoring, Xperia flags zero-tolerance events: abusive or dismissive tone, sarcasm, refusal to help, prolonged unresponsiveness, and inappropriate call drops. It distinguishes an agent hanging up from a customer hanging up, and does not penalise an agent for a transfer. IVR-only recordings with no agent are marked dead calls and excluded from scores.
Supervisors can override any category score, work code, summary or fatal flag. The original AI score stays retrievable afterwards, so a corrected call keeps an audit trail of what was changed and by whom.
See it working
Questions
Xperia scores 48 quality dimensions per call, grouped into 10 categories: Call Opening, Personalisation, Communication, Listening, Tone, Probing, Responsibility, Hold Procedure, Upselling and Closing. The set is configurable, so you can disable parameters that do not apply and add your own.
Yes. Category weightings are editable and must total 100, and scoring models are scoped per department, so different teams in the same account can be scored on entirely different criteria.
Yes. Any category score, work code, summary or fatal flag can be edited by an administrator or primary user. The original AI-generated score remains retrievable, so every override keeps an audit trail.
A fatal incident is a zero-tolerance breach that overrides the normal score — abusive or condescending tone, sarcasm, refusal to assist, prolonged unresponsiveness, or the agent dropping the call. Xperia distinguishes an agent hangup from a customer hangup and does not treat a legitimate transfer as fatal.
It can, because each custom parameter adds evaluation work to every call. Xperia calculates and shows the additional per-call cost before you save the parameter, so the impact on your per-minute rate is known in advance.
Related
Xperia's AI Assistant answers natural-language questions about your calls, agents and quality scores, returning tables and summaries instead of dashboards to navigate.
Score AI voice agents on 18 dimensions built for machines — barge-in handling, turn-taking, latency, hallucination, intent accuracy and context retention.
Every account starts with $3 of free credit — around 75 minutes of call analysis. No card required.
Start Free