Feature

A predictive dialer that paces itself on live answer rates.

Keep a floor of agents in live conversations instead of listening to ringtone. The pacing model reads answer rates, drop-rate limits and wrap times, and adjusts call by call.

How predictive dialing works

A predictive dialer places more calls than it has free agents. Most will not be answered by a person, so starting extra calls puts a live human on the line at roughly the moment an agent frees up.

The governing number is the pacing ratio. Too low and agents idle. Too high and someone answers to silence — an abandoned call. Everything that separates a good dialer from a bad one is how fast it re-reads that ratio, and what it refuses to do chasing occupancy.

Four dialing modes on one list

Campaigns trade agent occupancy against control of the conversation differently. All four modes run against the same list.

Predictive

Dials ahead of agent availability using live answer rates. Highest occupancy, best for large lists.

Progressive

One call per free agent. No abandoned calls by construction, at the cost of ring time.

Preview

The record opens first and the agent chooses when to dial. Right when history matters.

Manual

Agent-dialled, with the same recording, scripting and disposition rules applied.

Answering-machine detection

Machine pickups are classified before an agent is attached.

Voicemail drop

Leave a recorded message on a detected machine and move on, occupying nobody.

From record to conversation

01

The record is scored

Answer-rate prediction estimates a live answer from call history, time of day and number type. Unlikely records are re-ordered, not burned.

02

Suppression is checked at dial time

DNC and consent are evaluated as the call is placed. A number suppressed this morning is respected this afternoon.

03

The answer is classified

Ring-no-answer, busy, disconnected, fax, voicemail and live human are separated before anyone is connected.

04

The outcome is written back

Disposition, recording, transcript and summary go to the CRM record, with any follow-up task the disposition implies.

Pacing is a compliance problem, not just a throughput one

Abandoned calls are what regulators cap, which makes the drop rate a constraint to hold rather than a number to optimise against.

Firetone treats your ceiling as a hard rule. When conditions push toward the limit, the pacer slows. It does not run hot through your best hour and make it up overnight — that satisfies a monthly report while producing exactly the experience the rule exists to prevent.

Included with the outbound engine

  • Predictive, progressive, preview and manual dialing on one list
  • Answering-machine detection with optional voicemail drop
  • Drop-rate ceiling enforced continuously, not averaged
  • DNC and consent suppression checked at dial time
  • Dynamic caller ID by destination region
  • CRM write-back of disposition, recording, transcript and summary

The pacing model is inspectable

Every prediction lists the factors behind it, in the console and the export. Pacing, routing and handoff thresholds are manual controls you can pin. A model nobody can interrogate is a model nobody can defend to an auditor.

Predictive dialer questions

What is the difference between predictive and progressive dialing?

Progressive places one call per free agent, so nothing can be abandoned. Predictive places more, using live answer-rate data to predict how many connect.

Can I set a maximum abandoned-call rate?

Yes, and the pacer treats it as a hard rule rather than a target to average out. Pacing slows automatically as the projected rate approaches your ceiling, enforced continuously rather than reconciled at period end.

How does answering-machine detection work?

It classifies a pickup from the audio and timing of the greeting before an agent is attached. No detection is perfect, so the trade-off between certainty and speed is tunable per campaign.

What happens to numbers that do not answer?

They return to the campaign for retry, re-ordered by the best-time-to-call model rather than a fixed interval. Changing when you call usually beats calling more often.

See it running on your own traffic.

The free trial includes one module, ten seats and the full AI feature set, with an engineer configuring your first campaign alongside you.