Predictive vs power dialing: which mode fits your team

Fast answer: the vocabulary problem is a mapping problem

Choose the dialing behavior you need, then confirm the matching fields in your installed VICIdial build. In the inspected schema, MANUAL supports agent-initiated calls and preview options; RATIO uses a configured dial level; ADAPT_HARD_LIMIT, ADAPT_TAPERED and ADAPT_AVERAGE adjust pacing; SHARED_ variants support shared campaign staffing. The mapping below records the ten values in that inspected schema, not a promise that every release or customized Admin screen has the same options.

Power, progressive and predictive are product terms whose exact behavior varies between vendors. For this comparison, power means fixed-ratio automatic dialing and predictive means adaptive pacing. Ask any candidate vendor whether it starts one call per ready agent, multiple calls per agent, or calls before an agent is ready; the label alone does not answer that question.

The reader job is to choose a behavior and write a measurable trial plan. Record agent availability, live-person connections, time to an agent and call outcomes before changing the mode. Pacing fields describe configured intent; they do not establish measured abandonment or legal compliance.

MANUAL -> industry term: manual dialing. An agent selects or is served one lead; the dialer places exactly one call for that request.MANUAL with manual_preview_dial=PREVIEW_ONLY -> industry term: preview dialing. The agent sees the lead before it dials at all.MANUAL with manual_preview_dial=PREVIEW_AND_SKIP -> this column's own default. Preview with a skip-ahead option, still riding on the MANUAL dial method.RATIO -> industry terms: power dialing, progressive dialing. A fixed number of calls per available agent, set by auto_dial_level.ADAPT_HARD_LIMIT -> industry term: predictive dialing, with adaptive_maximum_level enforced as a hard ceiling on the dial level.ADAPT_TAPERED -> industry term: predictive dialing, tapered variant. Confirm its exact pacing behavior against your own build; the documentation available here does not fully specify it.ADAPT_AVERAGE -> industry term: predictive dialing, averaged variant. Same caveat: confirm against your build rather than assume a formula.INBOUND_MAN -> blended dialing: an agent can take inbound calls and place outbound manual calls in one logged-in session.SHARED_RATIO -> RATIO pacing shared across the agents of more than one campaign at once.SHARED_ADAPT_HARD_LIMIT -> ADAPT_HARD_LIMIT shared across more than one campaign.SHARED_ADAPT_TAPERED -> ADAPT_TAPERED shared across more than one campaign.SHARED_ADAPT_AVERAGE -> ADAPT_AVERAGE shared across more than one campaign.

MANUAL and preview: when a human decides every call

MANUAL makes the agent initiate each lead call. It avoids deliberate automatic dialing ahead of available agents, but a successful connection and timely agent handoff still need measurement. Preview adds a review step before dialing; it does not itself establish consent or compliance.

Preview dialing is not a separate dial_method value. It is a second, verified column on the same campaign row: manual_preview_dial, an enumerated field with three real values — DISABLED, PREVIEW_AND_SKIP, and PREVIEW_ONLY — defaulting to PREVIEW_AND_SKIP. Set it to PREVIEW_ONLY and an agent sees the lead's details before the call ever dials, has to accept it, and cannot skip ahead to another lead first. Set it to PREVIEW_AND_SKIP, this column's own default, and the agent keeps that same preview but can pass on a lead and move to the next one before dialing. Either way, the campaign's dial_method stays MANUAL underneath; preview is a layer on top of it, not a mode of its own.

This distinction matters the moment someone asks for a preview dialer. There is no dial_method called PREVIEW; there is MANUAL with manual_preview_dial turned to one of its two active values. Get that wrong in a request to whoever administers your campaigns and you will end up debugging the wrong field.

• MANUAL with manual_preview_dial=DISABLED: no preview screen, agent dials or is served a lead directly.
• MANUAL with manual_preview_dial=PREVIEW_ONLY: agent must review and accept before it dials.
• MANUAL with manual_preview_dial=PREVIEW_AND_SKIP: the same preview, plus the ability to skip a lead.

RATIO as power or progressive dialing: a fixed calls-per-agent number

RATIO holds a configured dial level instead of asking the adaptive worker to keep adjusting it. Read dial_method, auto_dial_level and available_only_ratio_tally together. A number without the tally rule and observed agent states does not describe the actual workload.

The inspected schema defaults are auto_dial_level=0 and available_only_ratio_tally=N. These are schema observations, not recommended production settings. Confirm your release’s tally behavior and current values before designing a trial; do not paste a universal RATIO=1.0 preset from this article.

A level of 1.0 is not evidence of zero abandonment. Calls may answer at different times, staffing changes during setup, and the agent handoff can still fail. Keep the initial authorized trial limited, agree on stop conditions, and compare measured live-person wait times and outcomes with the baseline before expanding.

SELECT campaign_id, campaign_name, dial_method, auto_dial_level, available_only_ratio_tallyFROM vicidial_campaignsWHERE campaign_id = '<CAMPAIGN_ID>';

The ADAPT family: what predictive dialing actually configures

ADAPT_HARD_LIMIT, ADAPT_TAPERED, and ADAPT_AVERAGE are the three values industry vocabulary collapses into predictive dialing. All three hand pacing to a background worker, AST_VDadapt.pl, instead of a ratio you set by hand; the ADAPT family adjusts dialing using answered, drop, and agent-availability observations plus campaign limits — that's what "adaptive" means here.

No source available here publishes the exact formula separating ADAPT_HARD_LIMIT from ADAPT_TAPERED and ADAPT_AVERAGE; the documentation available locally names what each ADAPT variant is for, not the precise pacing formula behind it. Confirm the exact behavior against your own build's release notes or by testing it, rather than trusting an assumed algorithm from this article or anywhere else.

What is verified, because each is a real vicidial_campaigns column with a real default, is the set of dials every ADAPT campaign shares: adaptive_dropped_percentage (defaults to 3, a percentage), adaptive_maximum_level (defaults to 3.0, a ceiling on the dial level), adaptive_latest_server_time (defaults to 2100, a four-digit cutoff time), adaptive_intensity (defaults to 0, on a scale where negative is less aggressive and positive is more aggressive), and adaptive_dl_diff_target (defaults to 0, weighing agents waiting for calls against calls waiting for agents). Every ADAPT_* value reads from this same set of knobs; naming the variant does not tell you where they are currently set.

SELECT campaign_id, dial_method, adaptive_intensity, adaptive_dl_diff_target,       adaptive_maximum_level, adaptive_dropped_percentage, adaptive_latest_server_timeFROM vicidial_campaignsWHERE campaign_id = '<CAMPAIGN_ID>';

SHARED_ modes: pacing across more than one campaign

The SHARED_ prefix on SHARED_RATIO, SHARED_ADAPT_HARD_LIMIT, SHARED_ADAPT_TAPERED, and SHARED_ADAPT_AVERAGE means exactly what it says: the same RATIO or ADAPT pacing logic, applied across agents who are eligible to take calls from more than one campaign at once, instead of one campaign's own dedicated agent pool.

Two things gate this in a real install. First, a system-wide System Setting named "Shared Agent Campaign Dialing" has to be enabled before a SHARED_ value does anything; until it is, the SHARED_ options still exist in the schema, but if your Admin Dial Method dropdown does not offer them, open Admin → System Settings and check that setting first — it is a global setting, not a per-campaign one. Second, once shared dialing is allowed, a verified vicidial_campaigns column, shared_dial_rank, defaults to 99 and lets you rank which campaign's calls take priority when agents are eligible for more than one.

Treat SHARED_ as a multi-campaign staffing decision, not a pacing decision. Whether SHARED_RATIO or a plain RATIO fits your team turns on whether you actually have agents working more than one campaign's queue at once; if you do not, the SHARED_ prefix adds a rank you have to manage for no benefit.

• Confirm the system-wide shared-dial setting is actually enabled before expecting SHARED_ options in Admin's Dial Method list.
• Use shared_dial_rank to decide which campaign wins when an agent is eligible for more than one.
• Do not adopt a SHARED_ mode unless agents genuinely work more than one campaign's queue.

Choosing by team size and call type, not by industry

Choose from a written requirement, then compare the same lead population and staffing conditions. Record whether agents must review the record before dialing, how unpredictable call duration is, the number of agents actually ready over time, and the reliability of live-person classification. There is no universal team-size threshold that makes adaptive pacing appropriate.

For a fictional team handling complex follow-up calls, agent review is mandatory and call duration varies widely. MANUAL with the required preview behavior is the first candidate because it satisfies that workflow. A second fictional team has repeatable short calls and a measured baseline; it can evaluate a controlled fixed-ratio or adaptive trial if its permissions, legal review and monitoring support it. Neither choice is justified by industry labels alone.

Change one factor at a time. Compare useful conversations per staffed hour with live-person wait time, operational drops and the applicable abandonment calculation. Stop or restore the prior settings if an agreed limit is crossed, classification becomes unreliable, or staffing falls outside the trial conditions.

• Write the required behavior, installed release and candidate field values.
• Save the current campaign values and baseline observation window.
• Name the trial owner, authorized population, staffing conditions and stop thresholds.
• Review the legal measurement population separately from VICIdial status totals.
• Accept the candidate only when the workflow and all agreed measurements pass; otherwise restore the recorded prior settings.

Measure abandonment separately from dialer settings

drop_call_seconds and drop_action control software behavior. Neither the field value nor a DROP status proves the timing of a person’s completed greeting and representative connection. This decision page does not recommend an eight-second/HANGUP recipe or treat a 1.0 ratio as a guarantee.

Under the U.S. TSR, an answered call is abandoned when a sales representative is not connected within two seconds of the person’s completed greeting. Its safe harbor requires technology limiting abandonment to no more than 3% of calls answered by people in each campaign’s defined period (the campaign if shorter than 30 days, otherwise successive 30-day periods or portions); at least 15 seconds or four rings before disconnecting unanswered calls; a prompt recorded seller name/telephone message when a representative is unavailable; and records establishing all those requirements. Other applicable federal/state rules and consent requirements still apply.

The FCC rule separately requires a telemarketing identification and opt-out message when a representative is unavailable within that two-second boundary. It includes an automated voice/key-press opt-out with instructions, automatic addition to the seller’s do-not-call list and immediate termination when selected. Its stated callback number must permit do-not-call requests during the campaign’s regular business hours and cannot carry prohibited premium charges. Review 47 CFR 64.1200(a)(6)–(7) alongside the TSR; a seller-name recording or HANGUP alone does not establish that these requirements are met. Neither abandonment provision grants permission to call a person without the required consent.

Have the responsible compliance owner define the applicable population, timing evidence, message and review period. Do not average campaigns together or divide abandoned calls by all attempts. Unknown human-answer classification is a measurement gap, not zero abandonment.

One campaign; one defined review period:1000 attempted calls200 calls answered by a person6 qualifying abandoned callsLive-answer abandonment = 6 / 200 x 100 = 3%6 / 1000 x 100 = 0.6% is an attempt-based ratio, not that measure.Ring-time, message and recordkeeping checks: separate and still required.

You should seeThe decision record separates software configuration, measured live-answer timing and the applicable legal review.

What to read next: changing the mode, then measuring it

This article stops at the decision. Once you know which dial_method, and, if relevant, which manual_preview_dial or adaptive tuning values fit your team, the mechanics of actually opening Admin, finding those fields, and raising a dial level in small measured steps are covered in VICIdial dialing modes and pacing: manual, ratio and adaptive.

After you switch modes, the only honest way to know whether the switch helped is to measure it: connect rate, drop rate, and how dial level interacts with answering-machine detection to create or destroy connected calls. That measurement layer, including the exact queries to run against vicidial_log and vicidial_carrier_log, is covered in Optimize VICIdial dial ratios and AMD for better connection rates.

Read this article to decide. Read those two to act, and to prove the decision was right on your own traffic, not on a mapping table.