Outbound calling stack — a staffed dial floor held to connect rate and cost per conversation

Most dial-floor stacks are assembled in the wrong order. The team buys a parallel dialer, sets it to the vendor’s suggested line count, watches dials per hour triple, and calls the project done. Two quarters later connect rates have fallen, half the outbound numbers carry a spam label, and nobody can produce an abandoned-call rate when legal asks for one.

This stack is built in the order the risk actually runs. The suppression layer decides which numbers get dialed. The dialer decides how many at a time — a number set by the abandonment cap, not by throughput. Gong records what happened on the connect so the floor improves instead of just dialing more. Apollo or Salesforce holds the record and, more expensively, the mobile numbers.

It is the human counterpart to the AI SDR stack and the AI outbound prospecting stack, both of which describe a motion with no rep on the line. This one assumes 5 to 25 people who dial for a living.

How the pieces fit

• The dialer is a swappable slot, and the three candidates are not interchangeable. Orum has the largest install base; Launch reports at $250 per user per month on an annual commit with a 3-seat minimum and up to 5 concurrent lines, and Ascend is the only product in the segment that reaches 10 lines. Koncert is the cheap-per-seat pick with coaching in the plan rather than as an add-on, and its parallel tiers cap at 5 lines — the “10X” on its pricing page is a productivity claim, not a line count. Nooks bundles coaching and account research around the dial at roughly $4,000 to $5,000 per user per year with a 5-seat minimum, which is a premium worth paying only when the salesfloor and scorecards get daily use.

• Gong is what separates a dial floor from a call center. Transcripts, talk-ratio and objection patterns turn 2,600 conversations per rep per year into a coaching curriculum. Without it, a parallel dialer raises the number of calls nobody listened to. Gong shipped MCP support on October 21, 2025, so an external agent can query call data without a middle tier you maintain.

• Apollo is the record and the number source, and the number source is the meter. Professional at $79 per seat per month annual includes the dialer and 100 mobile credits per month; Organization at $119 raises that to 200 with a 3-seat floor. A rep placing 300 dials a day exhausts a month of mobile credits in under an hour of list building. On a Salesforce-standardized team, Apollo supplies data and Salesforce stays the record — do not let a dialer or a data tool become the second place a disposition lives.

• The suppression layer is the component teams skip, and it carries the legal exposure. Two shapes exist. Pre-dial list scrubbing — DNC.com’s DNCScrub checks federal DNC, state lists via GeoScrub, internal do-not-call, wireless status via Wireless ID, and known TCPA plaintiffs via Litigator Insights, at up to 200,000 records per minute or a sub-second API call, with named integrations into 20-plus dialers. Or in-path blocking — Gryphon runs its own Tier-1 telephony network and drops a non-compliant call before it is placed rather than filtering the list beforehand. The first is cheaper and depends on your list staying fresh between scrubs; the second costs more and does not.

Named handoffs

1. List built → scrubbed → loaded. Every number leaving Apollo or Salesforce passes the suppression check before it reaches the dialer’s call list. Scrub at load time, not weekly: a number added to the federal registry on Tuesday is a violation on Wednesday.
2. Opt-out spoken → internal DNC → every future list. A rep hears “take me off your list” and the disposition writes to the internal suppression list, not just to the CRM activity. The FCC requires reasonable revocation requests to be honored no later than 10 business days after the request, and the internal list is the only mechanism that survives a list rebuild.
3. Connect bridged → Gong records → scorecard. The dialer hands the live answer to the rep and the recording starts. Consent notice fires here, before the first substantive sentence, not after.
4. Abandoned call logged → rolling rate → line count. Every burst where a second person answers produces a hang-up. That count divided by calls answered live, on a rolling 30-day window per campaign, is the number that sets the next week’s line count.
5. Spam label detected → number retired → pool rotated. Carrier analytics score high-volume bursts from a small number range. A flagged number loses most of its answer rate, so it leaves the pool before it drags the floor average, not after.

The line count is a compliance decision

This is the part vendors do not set for you. Under 47 CFR 64.1200(a)(7) a call is abandoned if it is not connected to a live sales representative within 2 seconds of the called person’s greeting, and no more than 3 percent of telemarketing calls answered live by a person can be abandoned, measured over 30 days per campaign. The safe harbor requires the abandoned call to play an identification and opt-out message — not silence.

Abandonment is a function of lines per burst and answer rate, and at a 5 percent answer rate it runs 2.5 percent at 2 lines, 4.9 percent at 3, 7.3 percent at 4 and 9.5 percent at 5. The 4-to-5-line configuration the dialers suggest lands at 2 to 3 times the federal cap. Better data makes it worse: verified mobile direct dials that lift answer rates to 10 percent roughly double abandonment at every line count. The full derivation is in parallel dialing.

Three rules follow, and they are the operating discipline of this stack:

• Run 2 lines on mobile-heavy lists and treat 3 as the ceiling. Above that you are choosing a rate you cannot defend.
• Make abandoned-call reporting a procurement blocker. If the dialer will not expose abandoned calls as a reportable field, you cannot measure the cap, and a vendor that cannot report it should not get the contract.
• Do not rest the argument on “we only call businesses.” The FTC’s Telemarketing Sales Rule carries the same 3 percent cap at 16 CFR 310.4(b)(1)(iv) and exempts business-to-business calls at 16 CFR 310.6(b)(7). The FCC rule has no equivalent carve-out in its text. The B2B answer satisfies one regulator and not the other.

Two adjacent facts belong in the same policy document. Under Facebook v. Duguid (2021) a dialer working an uploaded list is not an automatic telephone dialing system, which is why list-based parallel dialing avoids the consent-before-calling-a-mobile rule that governs true autodialers — but calls to a decision-maker’s personal cell remain subject to the TCPA’s wireless restrictions in a way that calls to a published business landline are not. Buying mobile direct dials is what moves a floor from the quieter lane into the regulated one. And the FCC’s revoke-all rule, which would extend one opt-out across unrelated campaigns from the same caller, was pushed again by a January 6, 2026 Bureau order to January 31, 2027; the core obligation to honor opt-outs is in force today regardless.

Recording carries its own map. Eleven states clearly require all-party consent — California, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, Nevada, New Hampshire, Pennsylvania and Washington — with Connecticut, Michigan, Oregon and Vermont unsettled enough to treat the same way. Interstate calls follow the stricter of the two states, and Gong’s own guidance is to obtain dual consent. For a floor dialing nationally, that resolves to one rule: announce the recording on every call.

Cost reality

For 10 reps on Orum Launch, per year:

The suppression range is the widest because the two shapes price differently. Published low-end anchors exist — TCPA Black List lists tiers at $80, $95 and $140 per month for 30,000, 50,000 and 100,000 monthly number checks, and $0.05 per check with no subscription. Gryphon and DNC.com are quote-only, and in-path blocking on a carrier network is the expensive end.

Now the number that decides the business case. At 3 lines, 2 dial hours a day, 120 dials an hour and 230 selling days, a rep places about 55,000 dials a year. At a 5 percent answer rate and 4.9 percent abandonment, that is roughly 2,600 conversations per rep, or 26,000 across 10 reps. The stack costs $2.15 to $3.80 per conversation. The reps cost $33 to $42 per conversation at an $85,000 to $110,000 fully loaded salary.

Tooling is 6 to 10 percent of the cost of a conversation. That is the whole argument: this stack has to raise conversations per rep to pay back, because there is not enough tool spend in it to save your way to a return. A team optimizing the dialer line item is solving the small half of the equation.

Common variations

• Koncert instead of Orum. Take it when seat price is the binding constraint and the coaching you need is live listen, whisper and a call library — those ship in-plan rather than as a paid add-on, and the entry dialer anchors near $150 per seat per month against a published half-price promo. The swap rule: if your throughput model needs more than 5 lines, this variation cannot serve it, and neither can Nooks.

• Nooks instead of Orum. Take it when a manager runs live coaching daily and the virtual salesfloor, AI scorecards and account research all get used. The floor is about $25,000 a year for 5 seats, so the rule is a coaching program that exists today — not one the tool is meant to create.

• Drop Gong; run the dialer’s native coaching. Take it under about 8 reps, where a manager can listen live and the curriculum fits in one person’s head. It removes $12,000 to $24,000 and the second recording consent surface. What you give up is searchable call history, which is the thing that makes onboarding rep 9 cheaper than onboarding rep 3.

• Trellus on top of an existing dialer. Take it when the dial stack is fine and the gap is in-call nudges. It is an add-on, not a replacement, and it does not change the abandonment math.

• Skip the third-party suppression layer. Defensible only when the list is entirely published business landlines, the dialer’s built-in DNC and time-zone guardrails are configured, and someone has written down why the mobile-dial exposure does not apply. Nooks documents built-in DNC and time-zone enforcement for the US, Canada and EMEA. Built-in guardrails do not cover litigator lists, and known-plaintiff filtering is the part that turns a nuisance claim into a settled one.

What this stack does NOT replace

• A list. Parallel dialing multiplies whatever the list quality is, in both directions. A stale list at 3 lines produces more dead air per hour and nothing else.
• A call script and an objection library. Gong shows you the objection that ended 40 percent of connects. Writing the answer is still a person’s job.
• A written calling policy. Line count, recording notice, internal DNC handling, and who reviews the abandonment rate monthly. If it is not written, it is not defensible.
• Legal review. State mini-TCPA statutes including Florida’s FTSA and Oklahoma’s OTSA carry their own definitions, and most but not all exempt B2B calls. This page is a stack design, not a compliance opinion.
• The follow-up motion. A connect that goes well becomes a meeting somebody has to run — see discovery call.
• An AI SDR. If the plan is to remove the human from the call, this is the wrong stack; read AI SDR first, because the volume-versus-rate trap in this page reappears there in email.

Match rules

Use this stack when:

• You staff 5 or more reps whose primary job is dialing, and the phone is a channel you intend to keep for at least a year. Under 5 reps the seat minimums on all three dialers waste the spend.
• Your buyer answers the phone. Field-service, logistics, healthcare operations and financial services still connect at 5 to 8 percent; some software-buyer personas do not answer at all, and no line count fixes that.
• Someone owns the abandonment number. A dial floor with no named owner for the rolling 30-day rate is running an unmeasured compliance exposure at increasing volume.
• Meetings per rep per week is already baselined. Parallel dialing raises dials per hour by construction, so the metric it is sold on always improves; the comparison has to sit somewhere else.

Do not use this stack when:

• The team dials fewer than 100 calls a rep per week. The single-line power dialer already inside Salesloft or Outreach covers that, at no additional contract.
• Your motion is inbound or product-led. Reps calling a scheduled inbound lead need speed to lead, not throughput — that is the inbound conversion stack.
• The list is mostly mobile numbers and nobody has run the abandonment math. Buy nothing until that number exists, because the better your data gets, the further out of compliance the default configuration puts you.
• Conversion per connect is the problem. Parallel dialing converts at roughly 60 percent of the per-connect rate of power dialing; if the connects you already get are not turning into meetings, more of them is the expensive answer to the wrong question.