Ask a room of collections supervisors what the call frequency rule says and you will hear "seven in seven" back within a second. Ask what happens on the eighth call and the room gets quieter, because the honest answer is it depends, and the rule was written to make it depend.
This is an orientation piece on what Regulation F's frequency provisions actually establish, what the word rebuttable does to them in both directions, and — the part that matters operationally — which system on your floor can actually prove your count. That last answer is probably not your QA programme, and it is worth knowing that before somebody asks you for the evidence.
Nothing here is legal advice. It stops where a legal conclusion would begin, which is the same place our coverage page stops.
What the rule actually established
The Fair Debt Collection Practices Act has prohibited causing a telephone to ring repeatedly with intent to annoy or harass since 1977. That language sits in the statute, public at congress.gov, and it is a standard rather than a number. For four decades the practical question — how many is too many? — had no published answer.
Regulation F, the Consumer Financial Protection Bureau's implementing regulation for the FDCPA, added one. The rule text lives in Part 1006 of Title 12 at the Electronic Code of Federal Regulations, the Bureau's own explanatory material sits at consumerfinance.gov, and the full preamble reasoning — which is where the interesting parts are — was published in the Federal Register.
Two presumptions came out of it, and they are frequently collapsed into one in conversation:
- A frequency presumption. More than seven telephone calls within seven consecutive days, in connection with a particular debt, is presumed to violate the harassment prohibition.
- A post-conversation presumption. A call within seven consecutive days after having had a telephone conversation with the consumer, in connection with a particular debt, is presumed to violate it as well.
The second one catches floors out far more often than the first, because it is not a volume rule at all. A single call can trip it. A collector has a productive conversation on Monday, the consumer says they will think about it, and a well-meaning follow-up on Thursday sits inside the presumption window even though the collector placed one call all week.
Rebuttable cuts both ways, and that is the whole design
The word doing the work is presumed. These are rebuttable presumptions rather than caps, and the preamble is explicit that the presumption operates in both directions.
| Below the threshold | Above the threshold | |
|---|---|---|
| What the presumption says | Presumed lawful | Presumed to violate |
| Can it be argued the other way? | Yes — conduct can still violate the harassment prohibition | Yes — the presumption can be rebutted on the facts |
| What decides it | The facts of the specific conduct | The facts of the specific conduct |
| What your record needs to show | Frequency and what was said | Frequency and what was said |
Read the bottom row twice. Whichever side of seven you are on, the evidence you would want is the same: the count, and the content. A floor that can produce a clean call count and nothing else has half of what the question asks for. So does a floor with a beautiful QA archive and no reliable per-debt count.
There is a second trap in the "particular debt" wording. The presumptions attach to a debt, not to a consumer and not to a phone number. One consumer with three placed accounts is three separate counts, and a system that tallies dials per phone number will produce a number that looks fine and answers a question nobody asked.
Which system on your floor owns the count
Here is the part vendors in our category tend to skate past, so we will say it plainly.
Call frequency is a question your dialer and your collections platform answer. It is not a question a call QA system answers, and ours does not. Frequency is a property of the account record — which debt, how many attempts, over which window — and that lives in the system that placed the calls and knew which account it was working. Our own stated limitation, which is on the features page and in the product itself, is that calls tie to collectors and times rather than to the account in your platform. Without an integration into whatever you dial from, we cannot count per debt, and a tool that guessed at that count would be worse than one that declines to.
So if frequency auditing is the problem in front of you, the first conversation is with your platform vendor, not with us.
What a complete QA record does answer is the other half — and it is the half that most often turns a defensible count into an indefensible one.
The things that only exist in the conversation
A dialer's counter sees attempts. It cannot see what was said, and several of the facts that change the frequency analysis exist only in speech:
- A cease request. "Stop calling me" is spoken on a call. If it never reaches your platform as a flag, the counter keeps counting happily and every later attempt is a different kind of problem than a frequency one.
- An inconvenient time or place. A consumer saying they cannot take calls at work, or before a certain hour, is a statement made in a conversation. The dialer records that a call connected, not that the consumer told your collector when not to call.
- Whether a conversation actually occurred. The post-conversation presumption turns on having had a telephone conversation. A ninety-second call that ended in a hangup is not obviously the same event as a real exchange, and the audio is the only place that distinction is visible.
- Which debt was actually discussed. A consumer with several accounts may have had one of them discussed on a call your system logged against another.
None of that is exotic. It is the ordinary content of a collections floor's day, and it is exactly the material a sampled QA programme mostly never sees — for the arithmetic reasons set out in why two percent sampling cannot tell you what your floor is doing.
The frequency argument is never really about the number. By the time anyone is arguing, both sides have the number. It is about whether we can show what the consumer said to us in week one and what we did about it in week two.
— Compliance director, ARM firm, 15 years, name withheld by request
What a floor can practically do this quarter
Four things, none of which require buying anything.
Establish which system is your system of record for the count. Write down the name of it. If two systems produce different counts — a dialer and a platform, or a platform and a spreadsheet — you have an evidentiary problem that will surface at the worst possible moment.
Check that your count is keyed to the debt. Not the consumer, not the phone number. If it is keyed to a phone number, find out what happens when two accounts share one.
Check how a spoken cease request travels. Trace the actual path: the collector hears it, and then what? A note field? A checkbox that stops the campaign? A supervisor who was meant to be told? Time how long that trip takes in practice. On most floors this is the weakest link in the whole chain and it costs nothing to look at.
Decide what your evidence pack looks like before you need one. For a single disputed account: the count, the attempt log, the audio, and the reviewed record of what was said. Assemble one for a random account this week as a drill. Whatever takes longest is the thing to fix.
The honest objections
"So your product does not help with the frequency rule at all?" It helps with one half and not the other, and we would rather say so than blur it. The count is your platform's job. The content of the conversation — a cease request, an inconvenient-time statement, a disputed debt, whether a real conversation happened at all — is what a complete call record gives you, with a timestamp and the quote attached. Buying a QA system expecting it to police your dial counts would be buying the wrong thing.
"Can automated review just detect cease requests for us?" It can flag language that looks like one, with the moment and the words, and that is genuinely useful because it surfaces the ones a collector did not register as a cease request at the time. It cannot decide whether a given phrase legally constituted one — that judgement belongs to a person who can be held responsible for it. The system decides what a human looks at. The human decides what it means.
"Our floor is well under seven, so this is not our issue." Possibly. But the post-conversation presumption is not a volume rule, and sitting under a threshold does not settle the harassment question on its own, because the presumption is rebuttable in that direction too. A low-volume floor with poor records is not in a better position than a high-volume floor with good ones.
"Does this differ by state?" State law can be more restrictive than the federal floor, and a number of states have their own debt collection statutes and licensing regimes. Our coverage page sets out what we do and do not assert about state rules, and the compliance page describes the attestation that gates recording in the software. Again: orientation, not advice, and your counsel owns the conclusion.
As of August 2026
As of August 2026 the presumptions described here are the operative federal framework, and the regulation text at the Electronic Code of Federal Regulations is the version to check rather than any summary — including this one. The Federal Trade Commission continues to bring debt collection enforcement actions and publishes its activity, which is a better read on what actually draws attention than any vendor's whitepaper.
Our own pricing, for completeness, is per monitored collector: $39 a seat for desktop activity and the morning report, $99 once every call is recorded, transcribed and scored, and $149 where screen recording and three-year retention are needed. Supervisors are free. The pricing page has the detail, and none of those tiers counts your dials for you.
Frequently asked questions
Does Regulation F cap collection calls at seven per week?
Not as a cap. It establishes a rebuttable presumption that more than seven calls in seven consecutive days, in connection with a particular debt, violates the harassment prohibition. A presumption can be rebutted on the facts, and conduct below the threshold can still violate the prohibition. Treating it as a hard ceiling is a reasonable operating posture, but it is not what the text says, and the distinction matters when you are explaining your programme to somebody.
What is the seven-day post-conversation rule?
A separate presumption: a call placed within seven consecutive days after having had a telephone conversation with the consumer about that debt is presumed to violate the prohibition. It is not a volume rule, which is why it catches out floors that are comfortably under the frequency threshold. One call can trip it.
Do the presumptions count per consumer or per debt?
Per particular debt. A consumer with several placed accounts produces several separate counts. Any tally keyed to a phone number rather than to a debt is answering a different question, and it will usually produce a friendlier-looking number than the real one.
Can a QA or call monitoring system prove our call frequency?
Ours cannot, and we say so directly. Frequency lives in the dialer and the collections platform because it is a property of the account record. What a complete call record adds is the content — cease requests, inconvenient-time statements, whether a real conversation occurred — which is the other half of the evidence, and usually the half that is missing.
Where should a floor look for the actual text?
Part 1006 of Title 12 at the Electronic Code of Federal Regulations for the regulation itself, the Consumer Financial Protection Bureau's published materials for its own explanation, and the Federal Register for the preamble to the final rule, which is where the reasoning behind the presumptions is set out at length. Summaries — again, this one included — drift.
What happens if a consumer asks us to stop calling?
That is governed by different provisions from the frequency presumptions, and the operational question is not really legal at all: it is whether the request reliably travels from the collector's ear into the system that decides who gets dialled tomorrow. Trace that path on your own floor and time it. Most of the failures we hear about in this area are failures of that handoff rather than failures of anybody's understanding of the rule.
Reviewing two percent of your calls?
CollectionsQA records, transcribes and scores every call your collectors take, and sends one supervisor report each morning naming the calls that need a human.