Home
Glossary

69 terms

The dialer glossary.

Every term a calling floor uses, defined the way it shows up at work. Regulatory terms link to the primary source. Operational guidance, not legal advice.

Dialing modes

#Auto dialer / ATDS

Any system that stores or produces numbers and dials them automatically; the TCPA calls this an automatic telephone dialing system.

An auto dialer is equipment that can store or produce telephone numbers using a random or sequential number generator and dial them, the definition Congress wrote into the TCPA as an automatic telephone dialing system, or ATDS. Courts and the FCC have argued for years about exactly what counts, and after Facebook v. Duguid the generator requirement does a lot of work. What matters operationally: calls made with an ATDS or an artificial or prerecorded voice to wireless numbers generally require prior express consent, and the penalties are per call. Know what your dialer is before you point it at cell phones.

#Click-to-call

A button in the CRM or dialer interface that places the call for the agent from the record on screen.

Click-to-call replaces hand dialing with a click: the agent opens a lead, presses the call button, and the system dials through the connected phone service while the record stays on screen. It removes typo retries, saves several seconds per call, and guarantees the number dialed matches the record. When it is embedded in the list view the agent never leaves the page, and every call is logged against the right lead automatically. It is the smallest upgrade over manual dialing and the first one most teams make. A power dialer goes one step further and places the next call without the click.

#Lines per agent

How many simultaneous call lines the dialer can run for each operator before answers outrun available humans.

Lines per agent is the practical dialing budget for one operator. One line per agent is the safe floor: every answered call has a person. Two or three lines per agent, as DialBreeze allows, raise the odds a given dial connects quickly because slow or unanswered lines stop wasting the agent's time; the tradeoff is that two live answers on one operator abandons someone. Above that you are running a small predictive system and should be watching abandonment math. Teams running multiple lines pair them with answering machine detection so machine answers do not burn the agent's slot.

#Manual dialing

An agent dials each number by hand from a list, with no automatic pacing at all.

Manual dialing is a person typing or copying numbers and pressing call, one record at a time. It is the slowest mode and the floor usually accepts it only when the list demands extreme care or the tooling offers nothing else: tiny high-value lists, sensitive collections, or numbers agents must not dial through software that could misfire. Every minute goes to lookup, dialing, and mistyped retries, so dials per hour is often half of what even a simple power dialer produces. Most teams keep manual dialing as a fallback, or skip it entirely once click-to-call exists.

#Pacing ratio

The number of lines a dialer places per available agent, the knob that trades abandonment risk against agent idle time.

Pacing ratio is lines per agent: at 2:1 a predictive dialer places two calls for every agent it expects to be free, at 3:1 three, and so on. Higher pacing fills agent time but multiplies the odds that two people answer when one agent is open, and every surplus answer is an abandoned call counting toward the 3 percent safe harbor limit. Low pacing keeps agents idle. Floors tune pacing from live answer rates, which shift by hour, list, and area code. Power and progressive modes sidestep the problem by never placing more calls than agents.

#Parallel dialer (multi-line dialing)

Dialing the same record on several lines at once and connecting the agent to whichever line answers first.

A parallel or multi-line dialer calls one record on multiple phone lines simultaneously, then bridges the agent to the first answer and hangs up on the rest. Each extra line raises connect speed and dials per hour, but every losing line that a human answers becomes an abandoned call, and machines answered on the other lines can waste even more time. Teams running multiple lines watch their abandonment rate closely and lean on answering machine detection to drop machine pickups fast. DialBreeze supports up to three lines per operator, enough parallelism to keep pace without a full predictive blast rig.

#Power dialer

A dialer that calls one number at a time per agent and moves to the next only after the agent finishes the previous call.

A power dialer places calls at a controlled pace: it dials the next number only when the agent closes the current one, so every answered call lands with a live human and nothing is abandoned. Compared with manual dialing, it removes dialing and lookup time; compared with predictive dialing, it never over-dials. On the floor, agents stay in rhythm: connect, talk, disposition, wrap up, next call. Power dialing suits lists where each contact matters more than raw volume. DialBreeze is a browser power dialer with a human on every call and support for up to three lines per operator.

#Predictive dialer

A dialer that places many calls per free agent, using statistics to predict answer rates, and drops the calls no agent can take.

A predictive dialer dials ahead: it calculates the odds that calls connect and places enough of them to keep every agent busy, hanging up on the people it mispredicts. Those dropped people are abandoned calls, which is why the TSR's 3 percent safe harbor exists. Floors run predictive dialers when lists are cheap and volume is the point: insurance blast campaigns, debt collection, large contact centers. The tradeoffs are real: pacing risk, abandoned humans, and no control over which contact an agent gets next. Regulated industries and consent-sensitive lists usually move to power or progressive modes instead.

#Preview dialer

A dialing mode that shows the agent the record first and dials only after the agent approves it.

In preview mode the agent sees the full record before any call is placed: name, history, notes, prior dispositions, and anything else that matters. The agent reviews, then clicks to dial. Nothing dials by surprise, which makes preview the mode of choice where context is expensive or mistakes are costly: regulated financial products, insurance renewals, legal services, high-value B2B. The cost is speed, since every record adds review time and dials per hour drops sharply. Some teams set a time limit on preview so agents cannot stall on records they would rather skip.

#Progressive dialer

A dialer that places one call per available agent and connects each answer directly, with no over-dialing.

A progressive dialer waits until an agent is free or about to be free, then places exactly one call for that agent. If someone answers, that person is connected; there is no gap and no abandonment, because no extra calls were placed. It sits between preview mode, where the agent approves each record, and power dialing, which keys the pace to the end of each call. Progressive pacing is common in collections, healthcare outreach, and any list where every answered call must reach a human with the right script in front of them.

Answer detection & voicemail

#Answering machine detection (AMD)

Software that listens after a call connects and decides whether a live person or a machine answered, routing each call accordingly.

AMD listens to the first seconds of audio after a pickup and classifies it as a live person, answering machine, fax, modem, or silence. On a calling floor this decides what happens next: a person is bridged to the agent, while a machine can be hung up or handed to a voicemail drop. Detection is never perfect, so teams tune sensitivity and accept a trade: aggressive settings catch more machines but occasionally hang up on slow-answering humans, which can count as an abandoned call under the Telemarketing Sales Rule. Well-tuned AMD keeps agents talking to people instead of listening to greetings.

#Beep detection

Detecting the exact moment a voicemail greeting finishes and the beep sounds, so a message or drop starts cleanly.

Beep detection is the timing problem inside voicemail handling: greetings vary from 4 seconds to 40, some ramble, some cut off, and the drop has to start right after the beep, not over the greeting or three seconds into silence. Good detection listens for the tone and the silence around it, then fires the recorded message. When it fails you get the classic artifact: half a message because it started late, or the recording talking over someone's outgoing message. Floors running drops check the failure rate by listening to sampled voicemails, and set a fallback of playing the message on any machine answer when detection is unreliable.

#Voicemail drop (ringless vs after-beep)

Leaving a pre-recorded message instead of a live one, either after the beep or ringless without the phone ringing.

A voicemail drop plays a pre-recorded message into voicemail so the agent does not repeat the same 20-second script five hundred times. Two kinds matter. A standard drop waits for the beep on an answered machine, then plays, keeping the agent moving to the next dial. A ringless voicemail drops the message into the mailbox without the phone ringing, using back-end carrier paths; its legal treatment under the TCPA is contested and depends on how courts read the definition of a call, so treat it as its own risk review, not a shortcut. Compliance aside, drops work: short, specific messages with a clear callback number get returned. In DialBreeze the operator leaves a prerecorded drop with one key when a line reaches voicemail.

Caller ID & carriers

#10DLC (10-digit long code SMS)

The registration system US carriers require for business texting from regular 10-digit numbers, applying to SMS rather than voice calls.

10DLC stands for 10-digit long code messaging: business SMS sent from standard local numbers instead of short codes. US wireless carriers require brands to register their campaigns through The Campaign Registry, with the business identified, use cases declared, and expected volumes approved, before traffic will flow at scale. Unregistered or mismatched traffic gets filtered hard. 10DLC matters to calling teams mostly because follow-up text often rides along with calls, and the same number handling both calls and compliant texts is cleaner for callbacks. It is a messaging rule, not a voice rule, but the registration ties to the same business identity and consent records.

#Attestation levels (A/B/C)

The A, B, and C grades an originating carrier assigns to a signed call, showing how well it knows the caller and the number.

Under SHAKEN, the carrier that originates a call signs it with an attestation level. Full attestation, A, means the carrier knows the customer and verified their right to use the calling number. Partial attestation, B, means the customer is known but the number relationship is not fully verified. Gateway attestation, C, means the call came from outside the carrier's direct knowledge, common with international or resale traffic. Receiving carriers use the level, along with analytics, to decide whether a call reaches the phone, gets labeled, or is blocked. Outbound teams buying service should ask what attestation their calls carry, because C-grade traffic converts poorly.

#BYOC (bring your own carrier)

Bring your own carrier: connecting the dialer to phone service you already have instead of buying it from the vendor.

BYOC means the dialing platform lets you plug in your own carrier account or SIP trunks rather than forcing you to buy per-minute service from the vendor. Teams want it for three reasons: pricing they already negotiated, numbers with reputation and registration history they have invested in, and the ability to keep everything if they change dialer vendors. The tradeoff is responsibility, since call quality, capacity, STIR/SHAKEN signing, and number hygiene move onto your carrier relationship, and the vendor's managed-service simplicity goes away. Floors with serious volume and existing carrier relationships generally prefer BYOC; small teams usually start on the vendor's built-in service. Twilio support in DialBreeze is planned.

#Caller ID

The number, and optionally name, displayed to the person receiving the call.

Caller ID is the phone number, and where CNAM is attached, the name shown to the person being called. For outbound teams it is both a trust signal and a compliance surface: under the Truth in Caller ID Act it is unlawful to transmit inaccurate or false caller ID information with intent to defraud or cause harm, and spoofing a neighbor number you have no right to use invites enforcement. Legitimate operations set accurate caller ID, keep business numbers registered, and use the same numbers consistently so callbacks work. When a number gets mislabeled, there are free remediation channels through the carriers and analytics providers, and teams should use them.

#Carrier / CPaaS

The telecom carrier or communications platform that actually places your calls and carries the audio across the phone network.

Behind every dialer is a carrier or CPaaS, a communications platform as a service, doing the actual telephony: placing calls, bridging audio, delivering caller ID, and interconnecting with the phone network. Twilio, Telnyx, Bandwidth, and Vonage are the names that come up, alongside traditional carriers. The provider determines call quality, STIR/SHAKEN signing and attestation, number inventory by area code, spam label treatment, and how easily your numbers stay clean. Floors choose on those grounds plus pricing per minute and per number. Swapping providers mid-program is painful, since numbers, reputation, and integrations all move, so the choice deserves diligence up front. Twilio support in DialBreeze is planned.

#CNAM

The 15-character name that shows up next to a phone number on the receiving phone's caller ID display.

CNAM is the calling name delivery service that puts a readable name next to a number on incoming calls. The name lives in databases the receiving carrier queries, and there are several of them, so updates propagate unevenly: a new number can show a blank, a stale name from a previous owner, or the carrier default. Outbound teams register their CNAM per number with each database that matters and fix stale entries, because a blank or wrong name reads as spam before anyone hears the script. Branded caller ID services push a name and logo through richer channels, but plain CNAM still does most of the work on ordinary phones.

#Latency and jitter

The two network measures that decide whether calls sound instant and clean or delayed and choppy.

Latency is the time audio takes to cross the network, and jitter is the variation in that time from packet to packet. Both are felt, not seen: high latency produces the awkward overlap where two people talk at once because nobody hears the other fast enough, while jitter produces robotic, choppy audio that makes prospects hang up. Voice wants latency under about 150 milliseconds each way and jitter in low single digits. On a calling floor the causes are usually local: wifi headsets, congested office networks, agents on home connections, or a provider route gone bad. Wired headsets, wired internet, and quality monitoring make more difference than any dialer setting.

Related: WebRTC, SIP trunk, Softphone

#Local presence dialing

Calling from a number whose area code matches the prospect's, so the call looks local and gets answered more often.

Local presence dialing presents a caller ID with an area code matching the person being called, because people answer numbers that look like their neighbors far more often than out-of-state ones. The mechanics are simple: a bank of numbers by area code, and the dialer picks one matching the prospect. The honest version uses real, registered numbers the business owns or rents and answers when called back; anything else decays fast, since unanswered callbacks kill the number's reputation and deceiving people with fake local numbers can violate caller ID rules. Teams using local presence should track answer rates and callbacks by area code and retire numbers that go bad.

#Number reputation

The standing a phone number has with carriers and analytics engines, which decides whether its calls ring, get labeled, or get blocked.

A number's reputation is the score carriers and analytics companies hold for it, built from call volume patterns, answer and decline rates, call duration, block and report counts, and whether the number is registered to a real business. Good reputation means calls ring through clean; bad reputation means labels, blocks, or silent filtering where the call never completes and nobody tells you. Reputation is why new numbers start slow and why burned pools stay burned. Managing it is operational hygiene: warm new numbers gradually, keep volume per number sane, monitor labels across the big analytics providers, and retire or repair numbers that drift.

#Number rotation

Spreading outbound calls across a pool of numbers so no single number burns out from overuse and labeling.

Number rotation dials through a pool of phone numbers instead of one, switching automatically to spread call volume. The point is survival: carriers and analytics engines label numbers as spam based on volume, short calls, and complaint signals, and one number carrying a whole floor's dials gets flagged. Rotation spreads the load so each number stays under the thresholds. It has costs: callbacks fragment across the pool, CNAM has to be managed per number, and rotating to evade labeling rather than to run a healthy campaign can itself be a red flag. Healthy rotation is wide, monitored per number, and paired with real answer rates, not throwaway burn-and-replace pools.

#SIP trunk

A voice service that delivers phone calls over the internet to a PBX or dialer instead of over copper lines.

A SIP trunk is a voice path over the internet: the carrier, or SIP provider, delivers calls to your system using the Session Initiation Protocol, and your dialer or PBX answers and places calls through it instead of physical phone lines. For a calling floor, SIP trunks are how dialing platforms get lines at scale: buy concurrent call paths, attach phone numbers, set caller ID per campaign, and route inbound callbacks back to agents. Quality depends on the provider's network and your internet connection, which is why latency, jitter, and packet loss are the three things to watch when calls start sounding choppy.

#Softphone

Software that turns a computer into a calling endpoint, with a headset instead of a physical desk phone.

A softphone is a phone that lives in software: a dial pad, call controls, and audio routed through the computer and a headset rather than a hardware phone. In dialing contexts the softphone is embedded in the agent's workspace, showing the record, script, and disposition panel around the call controls, so the whole interaction happens in one screen. Softphones cut hardware cost, move with the agent, and expose states a desk phone never had, like pause, wrap-up, coaching whisper, and hotkey dispositions. Quality comes down to the headset, the network, and the agent's machine, so floors standardize all three instead of letting agents improvise.

#Spam likely / call labeling

The label phones and carriers put on calls from numbers their analytics associate with spam or unwanted traffic.

Spam Likely, Scam Likely, and Telemarketer are labels that carriers and device analytics put on incoming calls from numbers with bad reputations. The signals are behavioral: high volume, short average call length, many unanswered or declined calls, and consumer reports. Once labeled, answer rates crater, because few people pick up a number the phone is already warning them about. Remediation is real: the major analytics providers run free registration and correction processes, and FCC rules push carriers to keep them available. The better long game is fixing the behavior that earned the label: longer calls, fewer dials per number, answered callbacks, and consistent CNAM.

#STIR/SHAKEN

The caller ID authentication framework that signs calls so receiving carriers can verify the number is real and belongs to the caller.

STIR/SHAKEN is the framework US voice providers use to authenticate caller ID on IP networks. The originating carrier cryptographically signs the call with the caller's identity, the terminating carrier checks the signature, and the call can be verified instead of arriving with a spoofed number and no trace. The FCC requires voice service providers to implement it under 47 CFR 64.6301. For outbound teams the practical effect is indirect but real: properly signed calls with accurate attestation are less likely to be blocked or labeled spam at the far end, while calls from non-compliant routes get filtered hard. Your dialer's carrier relationship determines how well your calls are signed.

#WebRTC

The browser technology that carries live audio in the web page itself, so agents can call from a tab with no desk phone.

WebRTC is the browser standard for real-time audio and video, the same technology video meeting apps use. In dialing it means the softphone is a web page: the agent opens the browser, signs in, and talks through a headset with no desk phone, no hardware phone, and usually no app to install. Audio flows from the browser to the platform's gateway, which bridges into the phone network through SIP or a carrier API. For distributed and remote teams this is the whole ballgame, since any computer with a headset becomes a calling station. Browser-based calling is what lets DialBreeze run a power dialer entirely in a tab.

Metrics

#Abandonment rate (abandoned call)

The share of answered calls where a live person is not connected to an agent within the required time.

A call is abandoned when a person picks up but no representative joins within two seconds of the person's completed greeting, per the FTC's Telemarketing Sales Rule. The abandonment rate is that count divided by all calls answered by a person, measured over a single campaign if it runs under 30 days, or over each successive 30-day period. The safe harbor caps it at 3 percent. Abandonment is caused by pacing: a dialer placing more calls than there are free agents. A human-on-every-call setup, where an operator hears every connection, effectively eliminates it.

#Appointment set rate

The share of conversations that end with a booked appointment or scheduled meeting, the standard front-end result for booked-sales teams.

Appointment set rate counts conversations that produced a scheduled appointment, the core output for teams whose closers run meetings, common in insurance, solar, home services, and financial services. It is measured against live conversations, not dials, because agents only set appointments with people they actually talk to. The rate exposes script quality and qualification discipline: low set rate with good talk time means conversations are happening but not converting, which is a script, offer, or targeting problem rather than a dialing one. Track set rate alongside show rate, since appointments that do not show are just expensive calendar entries.

#Connect rate

The share of dials that reach anything live, human or machine, versus ringing out or failing.

Connect rate is answered dials divided by total dials in a period. It is the health check on a list and a phone setup at once, because connects fail for both reasons: dead numbers, disconnected lines, and stale data on the list side; blocked calls, mislabeled numbers, and bad routes on the telephony side. A connect rate that drops while lists are unchanged points at number reputation or carrier trouble, not agent behavior. Teams track it by list source, area code, and hour of day, and act accordingly: scrub or replace the list, or repair and rotate the numbers.

#Contact rate

The share of leads in a list that were successfully reached at least once during a campaign.

Contact rate is broader than connect rate: of the leads on the list, how many were ever reached, live person or not, by the end of the campaign. A list of 1,000 with 400 contacts has a 40 percent contact rate, whatever the dials needed to get there. It measures list quality and persistence together, since most people need several attempts across days and times before they answer, and single-pass dialing leaves a large share of reachable leads untouched. Teams raise contact rate with cadence, varied calling windows, and multi-number records, and they watch for the floor where extra attempts stop producing new contacts, which is the signal to retire the list.

#Conversion rate

The share of contacts or conversations that produce the intended outcome: a sale, donation, appointment, or qualified meeting.

Conversion rate is outcomes divided by the base you choose: conversations, connects, dials, or leads. Pick the denominator and keep it fixed, because the same campaign can honestly report half a dozen different conversion rates. On a calling floor the honest chain runs dials to connects to conversations to conversions, and reading the chain tells you where money is lost: strong connects with no conversions indicts the script or offer; weak connects with strong conversion says the list is small but right. Conversion rate is also the metric that justifies everything else, since speed and volume only matter if something at the end of the funnel closes.

#Dials per hour

How many call attempts an agent or dialer completes per hour, the raw pace metric of an outbound floor.

Dials per hour is attempts divided by clock time, the simplest pace metric on the floor. What counts as a dial varies by tool, attempts placed versus numbers reached versus calls that rang, so define it once and stick to it. Pace is a function of mode: manual dialing lands far below click-to-call, power dialing adds pacing and wrap automation, and multi-line modes push higher still at the cost of abandonment risk. Read dials per hour next to connect rate and talk time, never alone: 120 dials with 3 connects is a list problem, not a productivity win. The number exists to be improved by removing dead time, not by hammering agents.

#Handle time

Total time an agent spends per contact: talk time plus hold plus after-call wrap-up.

Handle time is the full cost of one contact: ring and connect where counted, live talk, any hold, and all wrap-up until the agent is free again. It is the number capacity planning runs on, because agents available multiplied by handle time is the real throughput ceiling, whatever the dialer's theoretical pace. Handle time rises with offer complexity, required disclosures, and note-taking burden, and falls with hotkeys, templates, and AI summaries. Keep the components visible rather than merged: a floor with rising handle time needs to know whether it is longer conversations, which may be good, or longer wrap-up, which almost never is.

#Person-connect rate / right-party contact

The share of dials answered by a live human, and for sales work, the right person, sometimes called right-party contact.

Person connect rate counts dials that reached a live person, excluding machines, and in debt collection, insurance, and B2B the stricter form is right-party contact: the actual decision maker or account holder, not a spouse, gatekeeper, or wrong number. It is the metric that separates productive dials from noise, because a floor can post a fat connect rate on voicemail systems while never reaching a buyer. Improvements come from AMD tuning, list quality, calling windows matched to the audience, and local presence numbers. Track person connects per agent too, since some agents are simply better at getting past gatekeepers.

#Speed to lead

The elapsed time between a lead arriving and the first dial attempt, which collapses as the lead goes cold.

Speed to lead is the clock from lead creation to first contact attempt, and it is brutally predictive: contact odds fall off fast as minutes pass, and a lead called within five minutes answers at multiples of one called after an hour. Floors engineer it with automatic list ingestion, webhooks from forms into the dialer, and queue-jump rules that put fresh leads in front of the next free agent regardless of campaign order. The failure mode is structural, not personal: leads batch-imported at day's end, or sitting in an inbox until someone opens the CRM. Measure it as a distribution, and fix the slow tail, not the average.

#Talk time

The minutes actually spent in live conversation with a connected person, excluding ringing, dialing, and after-call work.

Talk time is the measured length of live conversation with a human, from bridge to hang-up. It is the numerator of a calling floor's existence, since only talk produces outcomes, and it is the first place to look when production drops. Compare it against dials per hour and connect rate: high dials with low talk means connects are failing or calls are ending fast, often a script or list problem; healthy talk with weak conversion points at the offer or the closer. Talk time per agent per hour, tracked as a share of login time, is the cleanest single measure of whether a floor is actually working.

#Wrap-up time

The after-call minutes an agent spends finishing notes, disposition, and next steps before becoming available again.

Wrap-up time, also called after-call work, is the gap between the call ending and the agent being ready for the next one: notes typed, disposition set, follow-up scheduled, sometimes a CRM field updated. It is the quiet thief of dialing capacity, because five minutes of wrap per call on forty calls is over three hours of non-talking. Floors attack it with hotkey dispositions, templated notes, and AI-drafted summaries the agent edits instead of writes, which is where after-call automation earns its keep. Watch the trend per agent: rising wrap-up usually means a process problem, a hard list, or an agent avoiding the phone.

Lists & workflow

#Attempt cap / max attempts

A per-lead limit on how many times the dialer will try a number before retiring it from the campaign.

An attempt cap, or max attempts, is the rule that stops a number being dialed forever: after N attempts without a connect, the lead drops out of the active rotation. Caps exist for economics and for sanity, since attempts six through ten rarely convert and hammering an unresponsive number burns the number's reputation and invites complaints. Good caps vary by lead type: inbound and opted-in leads deserve more attempts, cold data fewer, and the cap should interact with cadence, spreading attempts across days and times rather than firing all five in an afternoon. Watch contact rate by attempt number to find where your own list stops producing.

#Call cadence

The scheduled pattern of attempts for a lead: how many tries, spaced how far apart, across which days and times.

Cadence is the retry design of a campaign: how many attempts, at what intervals, at what times of day. A common pattern is a call plus follow-up on day one, calls at different hours on days two and three, then spacing of two or three days before the next attempt, retiring the lead after the cap. The logic is coverage: people answer at different hours, and repeating the same 10 a.m. slot five times is one attempt wearing five costumes. Cadence is set in the dialer per campaign and driven by dispositions, so no answer schedules the next attempt and callback requested schedules a specific time, automatically.

#Callback / follow-up task

A scheduled follow-up to call a specific lead back at a promised time, tracked as a task rather than a cadence step.

A callback task is a promise with a timer: the prospect said call me Thursday at four, and the system creates a task that fires then, in the right agent's queue if the floor works owned books. Callbacks convert far above cold dials because the person asked for the contact, so floors protect them: the task list drives the morning plan, and a missed callback is tracked like a missed appointment. Mechanically the dialer holds the record out of general cadence while the task is open, and closing the task with a disposition returns the lead to its normal flow or retires it.

#CRM sync

Keeping the dialer and CRM aligned: calls, dispositions, recordings, and notes written back to the lead records automatically.

CRM sync is the connective tissue between the dialer and the system of record: calls logged on the right lead, dispositions written back, recordings and summaries attached, statuses updated, and new inbound leads pulled in for dialing. It matters because the CRM is where the business keeps its memory, and unsynced dialing creates shadow data that reports miss and agents re-work. The two patterns are native sync, where the dialer writes through an integration, and webhook plumbing, where events fan out to wherever they are needed. The details that decide quality are field mapping, de-duplication on write, and whether callbacks and tasks created in the dialer appear in the CRM on time.

#CSV import

Bringing leads into the dialer by uploading a spreadsheet, with columns mapped to the fields the campaign uses.

CSV import is loading a spreadsheet of leads into the dialer: map the columns to fields, set the owner or campaign, and the records land in the queue. It is the entry point for most calling work, and its problems are the classics: phone numbers mangled by spreadsheet formatting, area codes split into their own column, dates in unusable formats, duplicate records inflating counts, and consent columns left unmapped so evidence disappears. A good import flow previews rows, validates numbers, flags duplicates, and requires a source and list name. Batch import is also why speed to lead suffers, since anything waiting for an end-of-day upload is aging exactly the way lead studies warn against.

#Disposition

The outcome code an agent assigns to a finished call, driving what happens to that lead next.

A disposition is the labeled result of a call: connected, no answer, voicemail, callback requested, not interested, wrong number, do not call, and so on. It is the most consequential field in a calling operation, because everything downstream keys off it: the cadence that decides when the lead gets dialed again, the stop-calling requests that must land on the DNC list, the reports that say which lists convert, and the agent stats that show who is working. Good disposition sets are short, mutually exclusive, and enforced at hang-up, so every call ends with one. Sloppy dispositions poison every metric the floor runs on, quietly and completely.

#Disposition hotkey

A single keyboard key that records the call outcome instantly at hang-up, no menus or mouse needed.

A disposition hotkey binds each outcome to one key: N for no answer, V for voicemail, C for callback, D for do not call. The agent presses it as the call ends and the system records the disposition, fires the next step in the cadence, and moves to the next call. This matters more than it sounds: when logging takes eight clicks, agents skip it or misclick it, and the data that drives cadence, compliance, and reporting rots. One key keeps the loop tight, a second per call instead of a minute. DialBreeze puts outcomes on number keys, and new dials wait until the wrap-up is saved.

#Lead list

The set of contact records a campaign dials, with phone numbers and whatever context travels with each one.

A lead list is the working set for a campaign: names, numbers, and usually source, status, time zone, consent flags, and history. Lists come from imports, purchased data, inbound forms, and CRM segments, and their quality sets the ceiling on everything downstream, because no dialer mode fixes numbers that are wrong or consent that is missing. Floors organize lists by source and age, since fresher data connects better, and by campaign, so each list can carry its own script, cadence, and attempt cap. List hygiene rules apply at the boundary: scrub against do-not-call registries and internal stops before the first dial, not after someone complains.

#Lead status

The stage label on a lead that tells the floor what has happened and what should happen next.

Lead status is the coarse state of a record: new, in progress, contacted, qualified, appointment set, closed, dead, do not call. It differs from disposition, which is per-call, in that status is the rollup, updated by the system or the agent as the relationship moves. Status is what routing runs on, since queues pull new leads for speed-to-lead treatment, hold in-progress leads to their cadence, and exclude dead and DNC records from dialing entirely. The discipline that matters is a short, owned status set with clear transitions. Floors with seventeen overlapping statuses get silent conflicts, where leads sit in states nobody works.

#List scrubbing

Cleaning a lead list before dialing: removing registry numbers, internal stops, duplicates, and bad or disconnected numbers.

Scrubbing is what happens to a list before it reaches the dialer: matching against the National Do Not Call Registry, honoring the internal do-not-call list and any state lists, dropping duplicates and disconnected numbers, and checking time zone and line type where it matters. It is a compliance gate and a productivity tool in one, since every dead number removed is a dial saved. The registry piece runs on a schedule, with lists re-scrubbed at least every 31 days to stay current with new registrations. Floors that skip scrubbing find out twice: first in connect rate, then in complaints and penalties, and the second one costs more.

#Webhook

A real-time HTTP notification your systems send or receive when an event happens, like a call ending or a lead arriving.

A webhook is a push notification between programs: an event fires, and one system POSTs the details to a URL another system owns. In dialing, webhooks move events the moment they happen: a call finished with its disposition, a voicemail dropped, a lead submitted a form, an appointment booked. They are the alternative to polling, where systems ask every few minutes and always lag, and they are what makes speed-to-lead engineering possible, since form to dialer can be seconds. Practical rules are boring and important: verify payloads, retry on failure, make handlers idempotent so a duplicate delivery does not double-book anything, and log everything during setup.

Recording & AI

#AI call summary

An automatic written recap of a finished call: what was said, what was asked, what happens next.

An AI call summary is the post-call write-up generated from the transcript: the reason for the call, what the prospect said, objections raised, commitments made, and the agreed next step. It attacks wrap-up time directly, since the agent edits a draft instead of composing notes from memory. The quality bar is specific: a summary that is fluent but wrong is worse than no summary, because the next person to touch the lead trusts it. Good setups constrain the model to the transcript, structure the output into fixed fields, outcome, next action, follow-up date, and keep the audio for verification. DialBreeze drafts a summary after each recorded call; missing or unusable audio can mean no summary.

#Call scoring

Grading calls against a rubric, by a reviewer or automatically, to measure script adherence and quality.

Call scoring grades a recorded call against a rubric: did the agent identify themselves, deliver required disclosures, follow the script, handle the objection, ask for the appointment. Traditionally a supervisor listened to a sample, maybe two calls per agent per week, and scored by hand. Automated scoring runs a model across every call, flagging the misses and ranking where a human should listen. The value is coverage and consistency, not judgment: rubric items that are objectively checkable, like disclosure presence, automate well, while nuance stays with the reviewer. Floors get the most from scoring when results tie to coaching within the week, not to a monthly report nobody reads.

#Call transcription

Turning the recorded audio of a call into written text, timed and labeled by speaker where the tool supports it.

Call transcription converts recorded audio into text, usually with speaker labels so you can see who said what. It is the substrate for everything downstream: searchable call history, QA scoring, AI summaries, sentiment reading, and dispute records. Accuracy varies with audio quality, accents, industry jargon, and cross-talk, so floors treat transcripts as reliable for search and review and verify anything that matters, like a verbal commitment or a compliance disclosure, against the recording itself. Transcripts also inherit the call's privacy weight: they contain everything said, so access control and retention rules should apply to the text exactly as to the audio.

#Conversation intelligence

The analytics layer over recorded calls: searchable transcripts, topic and question tracking, and patterns across thousands of conversations.

Conversation intelligence is the umbrella for analytics over call recordings at scale: transcribed, searchable, tagged by topic, question, competitor mention, objection, and outcome. It answers floor questions that used to require listening: which objections show up most on this list, what do the top closers ask that others do not, how often does the compliance disclosure actually get delivered, where in the script do prospects drop. The tooling matters less than the questions, since a searchable transcript library with honest outcome data answers most of them. The pattern to want is specific: conversation intelligence should change the script and the list, or it is decoration.

#Sentiment analysis

Estimating the emotional tone of a caller from their words and voice, per moment or across the whole call.

Sentiment analysis reads a call's tone: frustration, interest, confusion, anger, enthusiasm, estimated from word choice and, in some engines, from pitch and pace. On the floor it works best as a signal, not a verdict: flag the calls where sentiment dropped hard at the pricing line, or the ones that ended hot, and let a human decide what actually happened. Per-moment sentiment is more useful than a whole-call average, which mostly tells you the call was a call. The honest limits: people are polite while refusing, motivated while complaining, and models misread regional dialect and cross-talk. Use it to find the ten calls worth reviewing, not to grade the customer's mood.

#Speech-to-text

The underlying engine, often a speech model, that converts spoken audio into words, powering transcription and live agent aids.

Speech to text is the engine underneath transcription: an acoustic and language model that maps audio to words. The same engine increasingly runs live, not just after the call, which is what makes real-time agent assist possible: the model hears the prospect, surfaces the matching script section, flags a missing disclosure, or suggests an answer while the conversation is still going. Engine quality is measured by word error rate, and word error rate rises fast with noisy headsets, cheap audio paths, and domain vocabulary, so floors testing tools should score them on their own recorded calls, not vendor demo clips. Words matter here, because every downstream AI output inherits the mistakes.

#Whisper / coaching

A manager or trainer speaking into the agent's ear during a live call without the customer hearing.

Whisper coaching is the floor's live assist channel: the coach selects the agent, talks, and the agent hears it in one earpiece while the customer hears nothing. It is how new agents survive their first weeks, getting the next line fed mid-call, and how managers correct a call going sideways in real time instead of debriefing the wreckage afterward. The mechanics ride the conference bridge the platform already runs. Its close cousins are barge, where the coach joins audibly, and listen, silent monitoring; the three together are the standard supervision trio on calling floors. Use is a management style question: constant whispering builds dependence, occasional whispering builds skill.

See the terms in a working dialer.

Power dialing, voicemail drop, dispositions and after-call AI, in a 14-day sandbox.

Start a 14-day trial Read the guides