The two second test, the 3 percent safe harbor, and the records that make it stick: what US abandonment law actually says, written for the people who set dialer pacing.
What is the abandonment rate limit for predictive dialers in the US?
Federal rules cap abandoned calls at 3 percent of calls answered live by a person, measured separately for each calling campaign over each 30 day period, never against total dials. A call is abandoned if no agent takes it within two seconds of the person's completed greeting, and every abandoned call must receive a recorded identification message. The FTC's TSR and the FCC's TCPA rules both apply this standard.
The two second test, and why it starts later than you think.
The legal test is tighter than most dialer dashboards assume. Under the FTC's Telemarketing Sales Rule, an outbound call is abandoned if a person answers it and the telemarketer does not connect the call to a sales representative within two seconds of the person's completed greeting. That is the whole test, and it sits in 16 CFR 310.4(b)(1)(iv). The FCC's rules under the TCPA, at 47 CFR 64.1200(a)(7), apply the same two second standard.
Two details catch operators out. The clock starts at the end of the greeting, not the start of the call. "Hello?" takes a second; "Good afternoon, the Harrison household" takes four, and the rule does not let you average across them, so your transfer path has to survive the short greetings. And the definition covers every flavor of failure to connect: pure dead air, a drop after an awkward pause, and the one people miss, a live human your answering machine detection classified as a machine and silently discarded.
The baseline position of the TSR is blunt. Abandoning an outbound call is listed as an abusive telemarketing practice. Not some of them, any of them. A predictive dialer only operates legally because of the safe harbor, which tolerates a controlled level of abandonment provided every one of its conditions is met. Miss one condition and the shelter goes away entirely, not just for the calls over the line.
So the 3 percent figure everyone quotes is not a performance target with a tolerance band around it. It is the outer wall of a defense you have to be able to evidence, which is a different way of thinking about the number.
Per campaign, per 30 day period, against live answers. Each word is doing work.
Four conditions, all mandatory, no partial credit. The TSR safe harbor at 16 CFR 310.4(b)(4) requires technology that ensures abandonment of no more than 3 percent of all calls answered by a person, measured over the duration of a single calling campaign if it runs under 30 days, or separately over each successive 30 day period if it runs longer. It requires each call to ring for at least 15 seconds or four rings before the dialer gives up. It requires a recorded identification message whenever no agent is free within two seconds of the completed greeting. And it requires records proving all of it.
The FCC's text mirrors the arithmetic: no more than 3 percent "of all telemarketing calls that are answered live by a person, as measured over a 30-day period for a single calling campaign", with each successive 30 day period, or portion of one, calculated separately.
Read the denominator twice, because it is the most misquoted piece of dialer compliance we come across. The rate is abandoned calls divided by calls answered by a live person. Not dials placed. Not connects including voicemail. Live answers only. On a campaign that places 100,000 dials and gets 22,000 live answers (a worked example, not a benchmark), the 3 percent line sits at 660 abandoned calls, not 3,000. The machine pickups your detection correctly screened out never enter the calculation at all.
Per campaign matters just as much as the denominator. Each campaign carries its own measurement, so a new-data campaign running hot cannot be offset by a customer-base campaign running cold, and merging campaigns in your reporting layer does not merge them in the regulator's eyes. Each successive 30 day window of a long campaign gets its own calculation too, so a clean June does not purchase a sloppy July.
Our free 17 question Five9 Health Check covers pacing, AMD and campaign configuration, the places a regulator would start looking.
Measuring against dials, and measuring per day.
Myth one is measuring against dials. It survives because it flatters everyone: on a list where roughly one dial in five gets a live answer, the per-dial figure looks five times better than the legal one. Here is the worked campaign from above, calculated three ways with 900 abandons on the books.
| Calculation | Result | Standing |
|---|---|---|
| 900 abandons divided by 100,000 dials | 0.9 percent | Wrong denominator. Not the legal test under either regime. |
| 900 abandons divided by 22,000 live answers | 4.1 percent | The legal calculation. Over the line by a third. |
| Judging each day in isolation | Varies daily | Not the legal period since October 2008. |
Myth two is the per-day rule, and this one has a genuine historical excuse. The original standard really was daily, and the FTC amended the TSR in 2008 precisely because per-day measurement effectively banned predictive dialing on small lists, where a handful of abandons on a quiet day blows straight through 3 percent. Since October 1, 2008 the measurement runs per campaign over 30 day periods, and the FCC's rule matches.
The honest operator's footnote: we still track it daily on every floor we run, and you should too. Not because the law demands it, but because the 30 day window is unforgiving in the other direction. Two weeks at 5 percent needs the rest of the window near 1 percent to recover, and pacing a large team down to 1 percent abandonment while holding occupancy is miserable work. Daily tracking is how you avoid ever doing that math for real.
What has to play when no agent is there, and how long unanswered calls must ring.
When no agent is free within two seconds of the completed greeting, silence is not an option. The TSR requires a prompt recorded message stating the name and telephone number of the seller the call was placed for. The FCC's specification is stricter, and since most US outbound falls under both regimes, the FCC version is the one to build. Its message must play within two seconds of the completed greeting, is limited to disclosing that the call was for telemarketing purposes, must state the business name, and must give a no-charge callback number (not a 900 number) that takes do-not-call requests during business hours. It must also include an automated, interactive opt-out mechanism, voice or keypress, that records a do-not-call request and ends the call on the spot.
Notice the phrase "limited to". The message identifies; it does not sell. A message that slips in an offer stops being the abandoned-call cure and starts being a prerecorded telemarketing call, which under both regimes needs prior express written consent you almost certainly do not hold for that moment. We have audited setups where a well-meaning marketing team improved the abandon message and quietly converted a safe harbor condition into a fresh violation class.
The ring time floor rides along with the message: both regimes require an unanswered call to ring for at least 15 seconds or four rings before the dialer hangs up. Short-cycling unanswered calls to chase list velocity takes you outside the safe harbor even on calls nobody answered, which feels absurd until you remember the safe harbor is a package deal.
The FTC and FCC run parallel rulebooks that happen to agree on the arithmetic.
Operators use the acronyms interchangeably, and it costs them. The Telemarketing Sales Rule is an FTC regulation issued under the Telemarketing Act; the abandonment rules most people call "TCPA rules" are FCC regulations issued under the Telephone Consumer Protection Act. The two texts are deliberately aligned on abandonment, same 3 percent, same two second test, same 30 day campaign measurement, but they are separate rules with separate enforcers, and each reaches some callers the other does not.
| FTC regime (TSR) | FCC regime (TCPA rules) | |
|---|---|---|
| Rule text | 16 CFR 310.4 | 47 CFR 64.1200 |
| Abandonment standard | 3 percent of calls answered by a person, per campaign, per 30 day period | Same standard, same measurement |
| Abandon message | Seller name and telephone number | Adds the telemarketing purpose disclosure and an automated interactive opt-out |
| Enforcement | FTC, with DOJ filing civil penalty cases; state attorneys general | FCC forfeitures; state attorneys general; private TCPA litigation runs alongside |
| Exposure | Civil penalties up to $53,088 per violation (January 2025 adjustment) | Forfeitures, plus private statutory damages under the TCPA for related conduct |
The per-violation figure deserves a pause. The FTC's 2025 inflation adjustment puts the ceiling at $53,088 per TSR violation, the figure still in force at the time of writing, and courts have treated individual illegal calls as individual violations, which is how nine-figure judgments happen to companies that dial at scale. Private TCPA class actions, the cases that make trade press headlines, mostly run on consent and prerecorded-voice disputes rather than the abandonment standard itself; abandonment counts tend to arrive inside larger regulator complaints. That distribution is an observation from the public record, not a safety margin.
Rules in this space do not sit still, either. The TSR's recordkeeping section was rewritten as recently as 2024, and state statutes move yearly, so treat this article as a map: check the current regulation text before setting policy, and put anything with real money on it past qualified counsel.
Two cases, thirteen years apart, that bracket the range.
Numbers concentrate minds better than rule citations, so here are the two worth knowing.
The FTC's complaint charged the adjustable-bed seller with placing millions of abandoned calls: consumers answered to dead air, with no agent inside two seconds and no compliant message. Craftmatic paid a $4.4 million civil penalty, at the time the second largest in the do-not-call program's history, as part of a six-company sweep totaling roughly $7.7 million.
The government's long-running case against Dish covered tens of millions of illegal calls, mostly do-not-call and prerecorded violations by Dish and its retailers. A district court awarded $280 million in 2017; after an appeal vacated the figure, Dish settled in 2020 for $210 million, the $126 million federal share described by the DOJ as the largest civil penalty ever paid to resolve telemarketing violations.
One pattern stands out: pure abandonment cases are rare. Abandonment counts usually travel inside broader complaints about do-not-call and consent, because an operation careless about pacing is usually careless about list hygiene too. Which cuts both ways. Tight abandonment control will not save a dirty list, but it removes one count from any complaint, and it is the count most fully under the dialer team's control.
If you cannot show the math, you do not get the shelter.
Both texts are direct about this. The TSR's safe harbor conditions include retaining records establishing compliance with the other three conditions, and the FCC's rule states the seller or telemarketer must maintain records establishing compliance with its abandonment provisions. In an investigation the burden of showing the 3 percent discipline falls on you, and a dashboard screenshot from last quarter does not carry it.
The retention bar also moved recently. The FTC's 2024 TSR amendments extended record retention from two years to five and widened what must be kept, with the expanded recordkeeping in force from October 15, 2024. Five years is longer than most dialer platforms retain granular call detail by default, which quietly makes export and archiving a compliance function rather than an IT preference.
What we archive on the floors we manage, per campaign and per 30 day window:
Pacing discipline, AMD honesty, and someone watching the intraday number.
Compliance lives in three places on a predictive floor: the pacing engine, the answering machine detection settings, and whoever watches the intraday numbers.
Pacing first. Abandonment is the exhaust of over-dialing: the algorithm launched more calls than the agents coming free could take. Most abandonment breaches we investigate are not a pacing number set recklessly; they are a pacing number set reasonably for last month's list and then left alone. Fresh data answers more, so a pace tuned against a tired list will overshoot the day a new file drops, and midday shrinkage (breaks, a team huddle, one long wrap) does the rest. Operators typically hold an internal ceiling well under the legal 3, commonly around 2.5 percent, and treat the gap as budget for exactly these surprises. Our walkthrough of Five9 predictive dialer settings covers where the levers live.
AMD is the quiet saboteur. Every false machine detection is a live human who said hello, got silence, and was dropped: an abandoned call under the two second test, delivered without the required message, so it attacks two safe harbor conditions at once. Vendor accuracy claims are marketing until measured on your own traffic, and the honest answer is that every AMD configuration trades talk time against compliance risk. The trade should be a decision, not a default. We go deeper in our guide to Five9 answering machine detection.
Then monitoring. The 30 day window is the legal test, but the intraday per-campaign figure is the one that saves you: an alert at your internal ceiling, and a standing instruction to slow pacing rather than pause the campaign and hope. Across the floors we have rebuilt, nearly every meaningful reduction in abandon rate came from pacing and AMD discipline rather than anything exotic. One caution for transatlantic operations: if you also dial in the UK, Ofcom runs its own regime on different definitions, so do not port US math across unexamined.
Florida, Oklahoma and Washington change the consent question, not the arithmetic.
Federal rules are the floor, not the ceiling. Florida's Telephone Solicitation Act, passed in 2021 and amended in 2023, requires prior express written consent before using an automated system to select and dial numbers or play a recorded message to Florida residents, restricts calling to 8 a.m. to 8 p.m. local time, and caps attempts at three per 24 hours on the same subject. Oklahoma's 2022 statute tracks Florida's closely. Washington's law adds identification requirements within the first 30 seconds of a call and an obligation to end the call within 10 seconds of being asked. Florida and Oklahoma both carry private rights of action at $500 per violation, trebled to $1,500 for willful conduct, which is what turned Florida into a busy class action venue within months of passage.
None of these change the federal abandonment arithmetic. What they change is whether you can point a predictive dialer at a state's residents at all without written consent on file, and how expensive each individual mistake becomes once private plaintiffs can sue per call. Operators running national campaigns typically maintain state routing rules and suppress or re-consent the affected states rather than argue definitions later. These statutes are young and being amended in real time, Florida's was softened within two years of passage, so verify the current text before you build the routing table.
No. Both the FTC's Telemarketing Sales Rule and the FCC's TCPA rules define the rate as abandoned calls divided by calls answered by a live person. Unanswered dials and answering machine pickups never enter the denominator. On a list where one dial in five is answered live, the per-dial figure looks five times better than the legal one, which is exactly how dashboards that report abandonment per dial end up hiding a breach.
No. The FTC amended the Telemarketing Sales Rule effective October 1, 2008 to measure abandonment per calling campaign over each 30 day period, or over the campaign's duration if shorter, and the FCC rule matches. The old daily standard was dropped because it effectively banned predictive dialing on small lists. Experienced operators still track the figure daily, because recovering from a bad two weeks inside a 30 day window is brutally hard.
A call a person answers that is not connected to a live sales representative within two seconds of that person's completed greeting. The clock starts when the greeting ends, not when the call is answered. Dead air drops, late transfers, and live people wrongly classified as answering machines and discarded all count. The definition sits in 16 CFR 310.4 and is mirrored in the FCC's rules at 47 CFR 64.1200.
Under the TSR, the message must promptly state the name and telephone number of the seller the call was placed for. The FCC specification is stricter: the message is limited to identifying the business and disclosing that the call was for telemarketing purposes, must give a no-charge callback number that handles do-not-call requests, and must include an automated voice or keypress opt-out that records the request and ends the call. Most US operations build to the FCC version.
The substance is aligned: both cap abandonment at 3 percent of live answers per campaign per 30 day period, with the same two second test. The difference is ownership. The TSR is an FTC rule enforced through DOJ civil penalty cases, currently up to $53,088 per violation. The FCC enforces its own TCPA rules through forfeitures, states can act under both regimes, and private TCPA litigation runs alongside, mostly over consent and prerecorded messages.
Records establishing compliance with every safe harbor condition: per-campaign abandonment calculations with their live-answer denominators, call detail tying each abandoned call to its campaign and 30 day window, pacing configuration and change history, the deployed abandon message audio, and ring time settings. The FTC's 2024 amendments extended TSR record retention from two years to five, effective October 15, 2024, which is longer than many dialer platforms keep granular data by default.
Every false machine detection is a live person who answered, heard silence, and was dropped without an agent or the required message, which makes it an abandoned call that also breaks the message condition of the safe harbor. Vendor accuracy claims rarely survive contact with real traffic, so the practical approach is to measure false positive rates on your own campaigns and treat AMD sensitivity as a compliance decision rather than a throughput default.
Craftmatic paid a $4.4 million civil penalty in 2007 after the FTC charged it with placing millions of abandoned calls that left consumers with dead air. Dish Network settled the government's long-running telemarketing case in 2020 for $210 million, mostly over do-not-call and prerecorded violations, with the $126 million federal share described by the DOJ as the largest civil penalty ever paid to resolve telemarketing violations. The current FTC ceiling is $53,088 per violation.
The mini-TCPA statutes in Florida, Oklahoma and Washington leave the federal abandonment arithmetic alone and add stricter layers instead. Florida and Oklahoma require prior express written consent before automated dialing of their residents, cap attempts at three per 24 hours on the same subject, and carry private rights of action at $500 to $1,500 per call. Washington adds identification and hang-up timing requirements. The practical effect is state routing and suppression rules, not different abandonment math.
We build and run predictive floors under US and UK rules every day, as a dedicated pod rather than a ticket queue. If you want pacing, AMD and recordkeeping set up properly, or a second pair of eyes on what you have, talk to us.
Talk to us