Predictive or power dialer: which one your floor can actually carry.

What each one actually does

A power dialer places one call per agent. The rep finishes a call, the system dials the next number on the list, and the rep listens to it ring. If it rings out, they hear it ring out. If an answerphone picks up, they hear the greeting before they can hang up.

A predictive dialer places several calls at once, ahead of the agents, and only passes a call to a headset when a human has answered. The ringing, the engaged tones and the voicemail greetings all happen with nobody listening. The rep's next call starts at hello.

Two others sit alongside them and get confused with both:

• Preview dialling shows the rep the record first and lets them decide whether and when to dial. Slowest per call, and the right answer when the call is worth preparing for.
• Progressive dialling places one call per agent automatically, like a power dialer, but without giving the rep a chance to skip. It is a pacing decision rather than a different technology.

The maths that decides it

A predictive dialer works by prediction, and prediction needs a sample. It watches how many of your calls get answered, how long your calls last, and how long agents spend wrapping up, then dials enough numbers ahead to keep everyone talking.

With five agents there is not enough happening for that estimate to be stable. One long call swings the whole model, so the dialer either leaves people idle or connects calls with nobody free to take them. Those are your abandoned calls, and they are the ones regulators count.

This is the part most comparison pages skip, so here is the rough shape of it:

Those bands are not a standard. They are the point at which the statistics stop fighting you, and they move with your answer rate: a list that gets answered half the time behaves very differently from one that gets answered one time in twenty.

What each one costs you

Talk time is the honest measure. On a power dialer, a rep working a cold list can easily spend more of the hour listening to ringing and answerphones than talking. Predictive moves that waiting off the floor, which is the whole reason it exists.

The cost is on the other side. A predictive dialer will sometimes connect a call it has no agent for, and the person who answers hears silence and then a click. Every one of those is a real person you annoyed, and in most jurisdictions there is a cap on how often you are allowed to do it.

So the trade reads like this:

When a power dialer is the right answer

Reaching for predictive because it sounds more advanced is the most common mistake in this category. Power dialling is the better choice more often than vendors admit:

• You have fewer than about eight agents on the phone at once.
• The calls are worth preparing for. Anything where the rep should have read the record first is a preview or power job.
• You are calling people who already know you. Warm lists get answered far more often, which narrows predictive's advantage and raises the cost of dropping one.
• Your list is small enough that you will exhaust it. Predictive dialling wins by burning through volume, and there is no prize for finishing at eleven.
• You are in a regulated corner where any abandoned call is a problem worth avoiding entirely.

When predictive earns its place

Predictive is the right answer when the waiting is the cost, and that happens when three things are true at once: a large list, a low answer rate, and enough agents for the prediction to settle.

Collections, lead qualification, survey work, renewals chasing and most outbound BPO work all look like this. The list is long, most of it does not answer, and the value is in the small proportion that does. Every minute a rep spends listening to an unanswered phone is a minute not spent on the calls that matter.

If that is your shape, the question is not whether to use predictive. It is whether the system you buy will let you cap the damage, which is the next section.

The compliance part nobody enjoys

Predictive dialling is regulated because it can annoy people at scale. The specifics vary by where you are calling and change over time, so treat this as the shape of the problem rather than as advice, and take proper counsel for your own operation.

Three controls do most of the work, and any system you are considering should have all three:

• An abandonment cap. A hard ceiling on the proportion of answered calls that reach nobody, enforced by slowing the dialer down rather than by a report you read afterwards.
• Answering machine detection. Voicemail should never reach a headset, and just as importantly the detection should be tunable, because one tuned too aggressively hangs up on real people who answer slowly.
• Do-not-call screening on the way out. Screening at the moment of dialling, against your own suppression list as well as any register you are obliged to check.

The one that catches people is the abandonment cap being advisory. A number on a dashboard is not a cap. If the dialer does not actually slow itself down when it approaches the limit, you do not have one.

Questions to put to any vendor

Most of these have a bad answer that sounds fine until you ask the follow-up:

• Is the abandonment cap enforced by pacing, or is it a figure I read afterwards?
• Can I see the answering machine detection's false positive rate on my own calls, not a benchmark?
• What happens to pacing when three agents go to lunch at once?
• Is the call record created before the rep says hello, or written back afterwards by a sync job?
• Which of my numbers is used for a given call, and on what basis?
• Can I move a campaign from predictive to power without rebuilding it?