Call Center Glossary: 47 Terms Defined

Last updated: September 12, 2026

The four groups a call center vocabulary splits into

Call center terminology splits four ways. Operational metrics measure volume, speed, and staffing. Quality vocabulary describes how one call gets scored. Compliance vocabulary comes from statute and client contract rather than from operations. Technology vocabulary covers the stack that produces the recording and the transcript. Forty seven terms are defined below in that order.

Metrics that describe the queue and the call

Service level. The share of calls answered inside a target time, written as two numbers, so 80/20 means 80% answered within 20 seconds.
Average speed of answer. Mean wait in queue before an agent picks up, counted across a period instead of per call.
Abandon rate. Callers who hang up before reaching anybody, divided by calls offered. Short abandons under five seconds usually get excluded, and where that cutoff sits moves the number.
Hold time. Time on hold once the call is connected, separate from the queue wait.
Average handle time. Talk time plus hold plus the wrap that follows, divided by calls handled. AHT is the number a floor trades against quality.
After call work. Notes, dispositions, and dispatch that happen once the caller has hung up. Also called wrap.
First call resolution. The share of calls that settle the issue without a second contact about the same thing. Definitions run from same day to no repeat ever, so two centers both reporting 78% are often counting different windows.
Customer satisfaction. A post call survey score, usually the percentage picking the top one or two boxes. It records what the caller felt rather than what the agent did.
Net promoter score. Promoters minus detractors on the zero to ten recommend question, landing between minus 100 and 100.

None of those nine describes the review program itself. Which is not quite true of first call resolution, the outcome QA usually gets judged against, though it comes off the phone system rather than the scorecard. The numbers a program produces on its own sit on the QA metrics page.

The words a QA program runs on

Half of these change meaning between two floors in the same city, which is how a team inherits a scorecard and an argument together. They all belong to call center QA.

Scorecard. The written standard a call gets measured against: criteria, a point value on each, and the failures that void the whole thing, laid out on a 100 point QA scorecard template.
Criterion. One scored line, worded so a reviewer can mark it pass or fail without arguing about what it meant, held on a call monitoring form while the recording plays.
Weighting. The points a criterion carries, set out on a call scoring rubric. A card with 40 points on the greeting and 8 on the callback number read back digit by digit has told the floor what the company cares about.
Auto fail. A breach that takes the call to zero regardless of the rest. One floor calls it a fatal error and the next a zero tolerance item, and both words appear on the quality parameters page.
QA score. Points earned over points available on a call, averaged across an agent's scored calls.
Critical fail rate. Share of scored calls carrying at least one auto fail, counted rather than averaged, because a mean buries one skipped disclosure in a month of clean work.
Coverage rate. Calls scored divided by calls handled in the same window. Say an agent takes 1,400 calls a month and three get reviewed: coverage is around 0.2%, and everything the dashboard says about that agent rests on those three calls.
Calibration. A session where reviewers score the same call separately, then compare line by line. The spread between them is calibration variance, the number that says whether your scores mean anything. A calibration process keeps it from drifting.
Script adherence. Whether the required lines arrived, in the order the script sets, in the wording the client approved. An account script is the answering service version, one per client, and greeting scripts are the part most floors write down first.
Dead air. Stretches where neither side is talking, measured against a threshold such as ten or fifteen seconds.
Empathy statement. A line that acknowledges the caller's situation before the process starts. Machines score it unevenly, since a phrase match finds the sentence and cannot tell you whether the caller felt heard. Empathy statements has the phrasing.
Coaching. The conversation a score is supposed to produce, tracked as coaching completion, which is the whole argument for a coaching cadence.
Mystery shop. A staged call placed by a vendor and scored as if it were real, still the default in multifamily leasing. One shop a month per property is a sample of one, which is the case the mystery shopping alternative makes.

Compliance vocabulary

These come from a statute, a regulator, or a client contract, so the definition that counts is the one in the text rather than the one used on the floor. None of it is legal advice, and the version a supervisor can audit against is a call center compliance checklist.

One party consent. Federal law permits recording when a single participant to the call agrees, and your agent pressing record is that participant.
All party consent. Some state statutes require everyone on the line to be informed, or to agree. Published counts of those states disagree with each other, which call recording laws by state works through row by row.
Recording disclosure. The line at the top of the call saying it may be monitored or recorded. Fixed wording that either arrives or does not, so it scores as cleanly as anything on a card. The disclosure requirements cover where it is owed.
Retention schedule. How long recordings are kept before deletion, set by whichever rule on your list runs longest. Retention requirements puts them side by side.
Pause and resume. Stopping the recording while a caller reads out a card number, then restarting it, under the PCI rules about storing card verification data.
TCPA. The federal statute covering telemarketing calls, autodialers, prerecorded messages, and consent.
Do not call list. Numbers a company is barred from calling, both the federal registry and its own internal list. A request to be added, made on a live call, is a scoreable moment.
Mini Miranda. The FDCPA disclosure telling the consumer that the call is from a debt collector and that information will be used for that purpose.
TPMO. Third party marketing organization, the CMS term for a downstream entity that markets or enrolls Medicare Advantage and Part D beneficiaries. What a TPMO is covers who it catches.
The TPMO disclaimer. Required wording naming which plans the organization offers, delivered before plan benefits come up. Position is the part scorecards get wrong, which TPMO call compliance sets out.
Scope of Appointment. The beneficiary's agreement to discuss specific plan types, captured before the sales conversation starts. Scope of Appointment covers capture and storage.
Protected health information. Health data tied to an identifiable person. A service taking calls for a medical practice generally handles it as a business associate, which HIPAA call recording requirements works through.
Steering. Guiding a prospect toward or away from a property based on a protected class. It reaches phone wording long before policy, and fair housing phone compliance has the phrasing that gets flagged.

Technology vocabulary

Some of these describe the phone system and the rest describe the scorer, which matters mostly because it decides what a machine can check.

Automatic call distributor. The system that queues inbound calls and routes them by skill or account. ACD reports are where service level and abandon rate come from.
Interactive voice response. The menu that collects a digit or a spoken phrase before the call reaches a person.
Workforce management. Forecasting volume and building the schedule, including the on-call schedule that decides who covers dispatch at 2am. WFM sits next to QA and gets mistaken for it.
Automatic speech recognition. Turning speech into text. ASR is the layer everything downstream depends on, so its mistakes become the scorer's mistakes.
Diarization. Separating a transcript by speaker so each line lands on the agent or the caller. Mono recordings where both sides share one channel make it harder, and any criterion that depends on knowing who spoke gets weaker.
Dual channel recording. Audio with the agent on one channel and the caller on the other. It does more for scoring accuracy than any model change, and it is a setting on your recorder rather than a QA feature.
Confidence score. How sure the scorer is about one flag, carried beside it so a reviewer spends the hour on the shaky ones.
Keyword spotting. Matching a call against a stored phrase list, such as a required disclosure, a greeting string, or the do-not-say terms an account bans. A vocabulary library is where those lists live.
Sentiment analysis. A model's read on the tone of a call, reported as positive, neutral, or negative. Good for surfacing calls worth listening to and weak as a scored criterion, since the label and what the caller felt come apart.
Speech analytics. Search and trend analysis across a whole corpus of calls, answering what was said across thousands of them.
Conversation intelligence. Following an account across calls to find the patterns that precede a closed deal. Where these two stop and call scoring starts is the distinction buyers get wrong most often, which conversation intelligence against call QA settles.
Automated scoring. Running a scorecard against the transcript instead of a reviewer's ear. Rule based lines hold up well, including disclosure wording, a greeting string, and a dead air threshold. Tone stays weaker, which Aircall and StackAI both note, so AI call scoring is a division of labor.

Frequently asked questions

What is the difference between a call center and a contact center?

A call center handles phone calls. A contact center handles the phone plus email, chat, SMS, and whatever else the company answers. Everything above applies to both, though queue metrics get messier once one queue holds several channels.

What does QA stand for in a call center?

Quality assurance, the practice of reviewing recorded calls against a written standard and coaching from the results. QA analyst is the job title attached to it, and the scope sits in what call center QA is.

What is a good QA score?

SQM Group puts the industry average at 85%, calls 90 to 99% good, and reports that about 5% of agents reach 100. Those figures compare only across centers running similar scorecards, so your own baseline is the more useful target.

Why do two call centers report different numbers for the same metric?

Because the definitions underneath move. Abandon rate changes with the short abandon cutoff, first call resolution reads higher over seven days than over one, and a QA score means nothing without the card behind it.

Terms that mean different things depending on who is talking

Four groups is tidy right up to the point where one of these words crosses an industry line.

Parameter and criterion are the same column with two names. A QA analyst in Manila writes parameters at the top of the form and walks a client deck through them one at a time, while the identical line reads criteria once the program runs out of Phoenix. Quality parameters in BPO uses the first word for that reason.
First call. On a contact center dashboard it is the front half of first call resolution. At a funeral home the first call is the one that starts a removal, so an answering service holding that account gets scored on how it was dispatched at 3am against the on-call schedule in force. Answering service KPIs belong to the second job.
CSR. Customer service representative in the trades, where the person answering the phone turns a call into a booked job and gets graded on a CSR scorecard. Corporate social responsibility in a corporate report, which is most of what the acronym returns alone.
Fatal, auto fail, and zero tolerance name one mechanic across three floors. A demo using one word while your program uses another is how an imported scorecard quietly loses half its severity.
Monitoring. Live listening while the call happens in one shop, reviewing the recording afterward in another. CallSniper works on the second kind: audio arrives after the call ends, gets transcribed and diarized, then scored against the card you defined, on Trinsic's own model in Trinsic's own cloud.

Finding out which of these your own scorecard means takes about half an hour, so schedule a demo with a folder of calls you already have and watch them come back scored, at a fixed pilot price.

Sources

Transcription, diarization, vocabulary library phrase sets, confidence scores, and manual upload. CallSniper product documentation. Verified.
SQM Group on the 85% industry average QA score, the 90 to 99% band, and roughly 5% of agents reaching 100. Third party, unverified.
Aircall and StackAI on rule based criteria scoring more reliably than tone. Third party, unverified.
CMS rules for TPMOs at 42 CFR 422.2274 and protected health information at 45 CFR 164, carried from our compliance pages. Read from a primary text mirror.
Flagged for a check against current text: the FDCPA mini Miranda disclosure at 15 U.S.C. 1692e(11), the TCPA at 47 U.S.C. 227, PCI DSS on storing card verification data, and the Fair Housing Act as the basis for steering. Unverified.
First call as the funeral service term for the initial death call and the removal. Trade usage, unverified against a published glossary.
Parameter and criteria usage on BPO floors. Our own read, carried from the quality parameters page.