How we keep compliance.

Every rule, every enforcement point.

1. Calling-time window enforcement

The TCPA prohibits outbound calls before 8 AM or after 9 PM in the called party's local time zone. DialerSeat dials 9 AM to 9 PM, an hour inside the federal start. Before every dial, we resolve the lead to its time zone (from its state, or its area code when the state is missing) and check the current local time at that lead's location.

If the lead is currently outside its calling window, the dial doesn't fire. The skip is logged against the lead asTCPA_BLOCKED with the timestamp, the lead stays in the queue for when its window opens, and the campaign engine advances to the next eligible lead automatically. The agent doesn't have to think about it, and there's no setting to disable this.

Florida is stricter, and we hold Florida leads to Florida law: calls end at 8 PM their time, and no number gets more than three calls in 24 hours (Fla. Stat. 501.616(6)). A Florida lead that has had its three waits until the 24 hours are up.

2. Predictive abandon-rate monitoring

For predictive campaigns, the abandon rate is the number that matters. The FTC TSR caps abandoned calls at 3% of answered calls, measured on a rolling 30-day window. Cross that line and the safe harbor evaporates.

Our pacing controller computes each campaign's 30-day abandon rate from its own call records before every round it dials, and sets the number of lines from it: as many as keep the projected rate at or under 2%. The live rate is displayed on the agent terminal so nobody has to log into an admin panel to see where they stand.

3. Auto-degrade at 2.5%

The legal cap is 3%, but waiting until you're at 3% to intervene is too late. When the rolling rate hits 2.5%, the controller automatically degrades the campaign to 1× lines per agent (effectively progressive mode). One line can't abandon anyone. A line only comes back when the 30-day record, plus what that line could add, is projected to stay under 2.0%, and the campaign climbs back toward the configured multiplier as its record allows.

That 0.5% buffer is deliberate. It means a sudden answer- rate spike: caused by a bad list, a great script, or a statistical fluke, gets caught at 2.5% rather than at 3.1%. You should never see your campaign cross the legal threshold even under abnormal conditions.

4. The multi-line gate

Every extra line is a chance that two people answer at once, and only one of them can reach the agent. So a second line has to be earned. A new campaign dials one line per agent, which can't abandon anyone, and adds lines only when its own 30-day record can absorb a surplus answer and still be projected under 2%. The more people a campaign has already reached, the more lines it can carry.

Predictive also needs your business name and callback number, entered once in Settings (the dialer asks the first time you pick predictive), because the recorded notice below can't be played without them. Until both are set, predictive doesn't run. Agents pick 2 or 3 lines. When the dialer runs fewer lines than you asked for, it says which rule is holding it, so nobody is surprised.

5. AMD on progressive and predictive

Answering Machine Detection runs on every progressive and predictive call by default, via Telnyx's native call detection, with the same settings in both modes. Calls connect the moment they're answered, so a real person never waits in silence while a detector decides. When it hears a machine, the agent is released to the next lead and the call drops server-side. Machines never count as answered, which means the abandon-rate math isn't polluted by phantom "answered" calls that are just voicemail greetings.

The AMD result is stored on every call record. How AMD works has the full breakdown.

6. Ring duration ≥ 15 seconds

Per § 310.4(b)(4)(ii) of the TSR safe harbor, and the FCC's rule at 47 CFR 64.1200(a)(6), an unanswered call must ring for at least 15 seconds (or 4 rings) before being treated as no-answer. Our outbound calls are configured to honor this duration: predictive lines ring for 20 seconds, progressive for 30 to 43, preview and power for up to 60. A predictive line is never hung up while it rings, not even when you press stop.

7. Recorded notice on abandoned calls

When a person answers and their agent is already on a call, a teammate who is running predictive on the same team campaign and is free takes them instead. When an abandoned calldoes occur, meaning a human answered and no agent was available within 2 seconds, the TSR safe harbor requires a brief recorded notice with the seller's name and phone number, per § 310.4(b)(4)(iii), and the FCC adds that it must say the call is telemarketing and offer an automated opt-out. Ours names your business, says it's a telemarketing call, gives your callback number, and offers press 9 to be added to your do-not-call list. Pressing 9 adds the number and ends the call. The notice is built from the business name and callback number in your Settings (the team owner's, on a team campaign), with no generic fallback.

8. STIR/SHAKEN attestation

Outbound calls placed through DialerSeat go through Telnyx, which attests calls at the A level where the call path supports it. This is the carrier-level proof-of-identity that downstream carriers use to decide whether to label your call as spam, "Likely Scam," etc. A-level attestation doesn't eliminate spam-labeling risk, and attestation level can vary by number, route, or downstream carrier, it's the strongest level available, not a guarantee for every call.

9. Records: no expiry, exportable

Every dial attempt, AMD result, agent assignment, disposition, abandon event, and calling-window skip is stored with no expiry: nothing in the platform ages them out while your account is open. The TSR asks sellers to keep call records for five years (§ 310.5). Call records export to CSV from your dashboard's Compliance tab, a year at a time, for compliance audits, internal QA, or any other reason you need them.

Call recordings are the one exception: they're retained for 30 days, then automatically deleted (see our Privacy Policy). Download recordings before the 30-day window closes if you need to keep them longer: the surrounding call record (disposition, duration, AMD result, timestamps) stays in your dashboard regardless.

Three scenarios.

Scenario A: You upload a list at 11 AM ET

Your list contains 5,000 leads across all four U.S. time zones. When dialing starts, the system filters in real-time: leads currently in their local 9 AM-9 PM window are eligible, leads outside it are skipped with TCPA_BLOCKED and re- queued for when they enter their window. By 2 PM Pacific you'll have called a different mix than you did at 11 AM Eastern. The agent never thinks about it.

Scenario B: Predictive abandon rate spikes

You're running predictive at 2.0× with 6 agents. A campaign on a Tuesday afternoon hits an unusually high answer rate, maybe a fresh batch of warm leads dropped in and your in-flight calls outpace agents being free. The 30-day rolling abandon rate ticks up to 2.51%. Before its next round, the controller cuts you back to 1.0× lines. You keep dialing, but as progressive, not predictive. When the 30-day average works its way back down, far enough that a second line is projected to keep you under 2.0%, you resume 2.0× automatically.

Scenario C: You get an audit request

An attorney asks for records of every call placed to a specific phone number over the last 12 months. You export the campaign's calls from the Compliance tab to CSV, filter it to that number, and see every dial timestamp, AMD result, agent who took the call, disposition, and whether it was abandoned. Hand it over. You did the right things, you have the proof.