- AHT (average handle time)
- Mean time an agent spends on an interaction, including talk time, hold time and after-call work.
- BPO (business process outsourcing)
- A company that runs contact-centre operations on behalf of other businesses, usually several at once.
- Calibration
- The practice of having several reviewers score the same conversation and reconciling the differences.
- Campaign
- A batch of conversations grouped together, typically one client, queue or programme.
- Coaching plan
- The specific, evidenced feedback a supervisor gives an agent, based on what actually happened on their calls.
- Compliance score
- How completely a conversation met the regulatory and policy requirements that apply to it.
- Confidence level
- How often a sampling method would land within its stated margin of error, usually 95%.
- CSAT (customer satisfaction score)
- A direct rating from the customer, usually collected by a survey after the interaction.
- Dead air
- Silence during a live conversation, usually while the agent looks something up.
- Diarisation
- Determining who spoke when, by separating voices in the audio itself.
- Escalation risk
- An estimate of whether an interaction is heading for a complaint, a supervisor, or a churned customer.
- Evidence citation
- The specific lines of the conversation a score was based on.
- Fatal error
- A failure serious enough that the call fails regardless of how well everything else went.
- FCR (first contact resolution)
- The share of issues resolved without the customer having to come back.
- Inter-rater reliability
- How much two reviewers agree when scoring the same conversation.
- Margin of error
- How far a result from a sample might sit from the truth about the whole population.
- Model confidence
- How sure a model reports being about a judgement it just made, not a guarantee that the judgement is right.
- NPS (net promoter score)
- A single-question loyalty measure (how likely you are to recommend), reported as promoters minus detractors.
- PII masking
- Removing personally identifying details from a transcript so the record can be reviewed without exposing them.
- QA scorecard
- The list of criteria a conversation is scored against, each with a weight and a severity.
- Quality assurance (QA)
- Reviewing customer conversations against a defined standard, so that performance can be measured and coached rather than guessed at.
- Rate limiting
- Deliberately capping how fast a batch is processed.
- Root cause analysis
- Working out why a call went wrong, rather than only recording that it did.
- Sentiment analysis
- Estimating the emotional tone of a conversation, usually for each speaker.
- SOP (standard operating procedure)
- The written process an agent is meant to follow for a given type of interaction.
- Statistical sampling
- Choosing how many conversations to review so that the result says something reliable about all of them.
- Talk-to-listen ratio
- How much of the conversation the agent spoke versus the customer.
- Transcript
- The written record of a conversation, usually with each turn attributed to a speaker.
- Weighting
- How much each criterion contributes to the overall score.
- ZTP (zero tolerance policy)
- A category of failure with no acceptable threshold at all, usually regulatory or safety-related.