Predictive, power, or preview: which auto dialer mode to choose

With manual dialing, an outbound agent spends less than half the workday actually talking to customers. The rest goes to keying in numbers, ringing, busy signals, and voicemail greetings – and you pay for that time too. An auto dialer eliminates this waste, but its three modes work very differently. Pick the wrong one, and either your agents sit idle or customers start getting calls where nobody's on the line. Let's compare the modes and calculate what a conversation costs in each.

Preview dialing: the agent decides when to call

The system shows the agent the next customer's record (preview). The agent reads it and places the call when ready, or skips the record.

• Good for complex B2B sales, high-value customers, debt collection that requires preparation, and callbacks on complaints.
• Pros: no abandoned calls, and the agent knows who they're calling and why.
• Cons: the fewest calls per hour – the agent waits for every answer.
• Tip: cap the time spent reviewing a record, or preparation turns into idle time.

Power dialing: one number per available agent

As soon as an agent becomes available, the system dials the next number (power dialer). The customer who picks up hears the agent right away. Some setups dial two lines per agent – that's faster, but it introduces abandoned calls.

• Good for mid-size lists, warm contacts, reminders, and service calls.
• Pros: no manual dialing, and with one line per agent there are almost no abandoned calls. Small teams can handle it too.
• Cons: the agent waits for the customer to answer and loses time on no-answers unless answering machine detection filters them out.
• Tip: a good first step for a team moving away from manual dialing.

Predictive dialing: dialing ahead based on statistics

A predictive dialer calculates how many numbers to dial so that a live customer answers right when an agent becomes free. It factors in what share of customers pick up, how long calls last, and how many agents are on the line. If one in five people answers, a power dialer makes five dials on average while the agent waits for a live customer. A predictive dialer dials those numbers in advance and in parallel.

• Good for large cold lists where few people answer, high-volume campaigns, surveys, and short sales calls.
• Pros: agents are almost never idle and spend the most time in conversation.
• Cons: if more customers answer than there are free agents, some calls get dropped or the customer hears silence.
• Tip: you need at least 30 agents on one campaign – we explain why in the sidebar below.

Why predictive mode needs at least 30 agents. Dialing has to start about 25 seconds before an agent becomes free – that's how long it rings before someone answers. So the system has to guess how many agents will free up in the next 25 seconds. If a call plus after-call work takes 3.5 minutes, with 30 agents someone frees up every 7 seconds on average: that's 3-4 people in 25 seconds, and forecast errors cancel each other out. With 5 agents, one frees up every 42 seconds, and in the same 25 seconds either no one or one person frees up. Every error means a customer with no one to answer or an agent left waiting.

What one conversation costs in each mode

The difference between modes is the money you pay agents to wait. Let's run an example. One in five people answers. An unsuccessful attempt – ringing, busy, voicemail – takes about 25 seconds. The conversation lasts 3 minutes, plus 30 seconds for the agent to log the outcome. An agent hour costs $50.

• Manual dialing: for each conversation, the agent dials five times at 15 seconds each and waits in vain four times. The cycle is 6.4 minutes, 47% of the time is spent talking, about 9 conversations an hour. Each conversation costs $5.33.
• Power: the system dials, but the wait while it rings remains. The cycle is 5.2 minutes, 58% of the time is spent talking, 11-12 conversations an hour. Each conversation costs $4.33.
• Predictive: while the agent is talking, the system is already dialing the next numbers. About 10 seconds remain between conversations, 82% of the time is spent talking, 16 conversations an hour. Each conversation costs $3.13.

For a 10,000-conversation campaign, that's about $53,300 of agent time with manual dialing, $43,300 in power mode, and $31,300 in predictive mode. A team of 30 has about 2,800 conversations in power mode over an eight-hour day and about 3,900 in predictive mode. But predictive has its own cost: even at a 3% abandonment rate, about 120 people a day hear silence instead of an agent, and some of them won't pick up again. That's why with a small team, high-value customers, or a strict abandonment limit, power or preview dialing often delivers a better bottom line.

Answering machine detection

A large share of outbound calls reach an answering machine, voicemail, or a virtual receptionist. Answering machine detection (AMD) recognizes this from the first seconds of the answer and either ends the call or schedules a new attempt, so agents talk only to people. In predictive mode, it's a must: without it, the dialing pace is calculated wrong.

• Balance speed and accuracy: fast detection shortens the pause for a live customer but makes more mistakes.
• Watch for false positives: a live person mistaken for a machine is a lost contact. Regularly listen to a sample of calls flagged as answering machines.
• Tune detection on your own lists: greetings vary by carrier and region. Perfect accuracy doesn't exist.
• Redial numbers that hit an answering machine at a different time of day.

What US rules require

Outbound calling in the US is regulated at the federal and state level. Set up your dialer to enforce the rules automatically instead of relying on agent discipline. Check with your lawyer which rules apply to your campaign.

• Caller ID. Call from numbers registered to your company. Caller ID authentication (STIR/SHAKEN) is handled by your carrier; calls that can't be verified are more likely to be flagged as spam or blocked.
• Do Not Call. Customers can add their number to the National Do Not Call Registry or ask your company to stop calling them. Load your Do Not Call lists into the dialer so those numbers are never dialed.
• Consent. Under the TCPA, autodialed and prerecorded telemarketing calls require the person's prior consent, and you must stop calling as soon as they ask. Don't send prerecorded sales messages without a live person unless you have that consent.
• Personal data. A call list is personal data. Follow CCPA/CPRA and other state privacy laws when you collect and use it.
• Debt collection. Set rules for collection calls that keep you within the law: allowed hours in the customer's time zone and limits on how often you call.

For telemarketing calls, US federal rules limit the abandonment rate to 3% of answered calls per campaign over 30 days and require a short recorded message when no agent is available – but whatever applies to you, keep the rate as low as possible: a customer who hears silence often won't pick up again. Set a limit in your campaign settings, play a short message if no agent is free, cap the number of attempts per number per day and per campaign, and call only during hours that suit the customer's time zone (under the TCPA, 8:00 AM-9:00 PM local time).

Preparing the list and attempts

The mode is only half the story. The other half is how the list is prepared and when each number gets dialed.

• Clean the list before the campaign: remove duplicates, invalid numbers, and customers who opted out of calls.
• Segment the list by customer value, region, and time zone, and assign each segment its own mode.
• Spread attempts across different days and times: someone who doesn't answer in the morning may answer in the evening.
• Stop dialing numbers that are unreachable or go to voicemail time after time – try another channel.
• Let agents schedule a personal callback if a customer asks to be called at a specific time.
• Show the agent the customer record and a short script as soon as the call connects.

A well-prepared list in power mode often outperforms a raw list in predictive mode: agents spend their time on the right people at the right moment.

Which metrics to track

• Talk time as a share of agent working time.
• Conversations per hour and conversion per conversation.
• Abandonment rate by campaign and hour.
• Answering machine detection accuracy.
• List penetration: what share of numbers you reached after all attempts.

Where to start

1. Assess your list: its size, the share of customers who pick up, and the value of each contact.
2. Count how many agents work the campaign at the same time. If it's fewer than 30, choose between power and preview dialing.
3. Check which rules apply to your calls.
4. Start with power dialing, and after a week calculate the cost per conversation: agent hours × hourly cost ÷ number of conversations.
5. Try predictive mode on part of the list and compare cost per conversation, abandonment rate, and conversion.
6. Keep preview dialing for customers where preparation noticeably improves conversion.

You can mix modes: the mode is chosen per campaign. For example, run a large cold list in predictive mode, warm contacts in power mode, and key accounts in preview mode.

The CONTACT.WORK auto dialer runs in preview, power, and predictive modes – with answering machine detection, an abandonment rate limit, and rules for attempts and intervals. The list comes from your CRM through the API, results are written back, and reporting on outbound campaigns and inbound queues lives on one platform.