Two of the most consequential things a consumer can say to a collector are said out loud, in passing, in the middle of a call. Stop calling me. I don't owe this. Neither arrives as a letter, neither announces itself, and both change what happens next.
Whether the floor then behaves correctly depends on something mundane: did the collector notice, and did the right box get ticked in the platform afterwards? That is a human step in the middle of two systems, and it is exactly the kind of step that fails quietly and stays failed until a complaint arrives.
This is orientation, not legal advice. It describes where the seam is and what a record needs to be able to show. Where a genuine legal conclusion begins — whether a particular sentence counted, what your obligation was in a particular case — that is your counsel's question and we stop there deliberately.
What the rules put in play
The Fair Debt Collection Practices Act has been on the books since 1977 and its consumer-facing mechanics sit in two places worth knowing by name.
Cease communication. The statute provides that where a consumer notifies a collector in writing that they refuse to pay or want communication to stop, the collector must not communicate further except in narrow, specified ways — such as advising that collection efforts are being terminated or that a particular remedy may be invoked. The statutory text is public at Congress's own site and in the Electronic Code of Federal Regulations.
Disputes and validation. The Act also gives the consumer a period in which to dispute the debt, and provides that where a dispute is made in the specified manner the collector ceases collection until verification is obtained and mailed. Regulation F, which implements the Act, was finalised by the Consumer Financial Protection Bureau and its rulemaking record is at the Federal Register; the Bureau's own explanatory materials sit at consumerfinance.gov.
There is a formality question here that people flatten in conversation and should not. Some of these provisions are framed around writing; some routes are electronic; state law and creditor contracts frequently add obligations of their own, and many agencies adopt an internal policy that treats an oral request as actionable regardless of what the federal minimum is. Which of those applies to you is not a question this page can answer, and any vendor telling you otherwise on a marketing page is selling you a risk.
What we can say without qualification is operational: if a consumer said something on a call that could be read as a cease request or a dispute, and nobody on your side can find that moment eight months later, then whatever the legal answer turns out to be, you cannot show what happened. That is a records problem, and records problems are ours.
The seam, drawn precisely
Follow the sentence from the consumer's mouth to your system of record and count the places it can fall through.
| Step | What has to happen | How it fails |
|---|---|---|
| The consumer says it | Plain or oblique — "quit calling", "this isn't mine", "talk to my lawyer" | Said mid-sentence, over the collector, or as an aside |
| The collector notices | Recognises it as significant, not as an objection to work past | Heard as a stall; training gap; call momentum |
| The collector responds correctly | Follows your script for the branch | Continues the pitch; asks them to put it in writing badly |
| The platform is updated | Correct status set on the correct account | Set on the wrong account; set late; not set |
| The suppression takes effect | Dialer and letters stop as policy requires | Campaign already loaded; letter already queued |
| It can be proved later | Somebody retrieves the moment on request | Recording unindexed; nobody knows which call |
Only the first and last rows are outside the collector's head. Everything between them depends on one person recognising one sentence while under pressure to collect — and the platform side of the record, which is what most audits examine, only ever shows the result of that recognition. A clean dispute log is equally consistent with a floor that catches everything and a floor that catches half and never notices the other half. The log cannot tell those apart. Only the conversations can.
This is the same structural point as the seven-in-seven frequency presumption: the platform owns the count, but the things that determine whether the count means anything happen in the conversation, where the platform is blind.
What a QA record has to be able to show
Set aside detection technology for a moment. Independently of how you get there, a record that is any use in this area answers five questions:
- When was it said? Not which call — which second. A finding that points at a twelve-minute recording is a finding somebody has to redo.
- What exactly was said? The consumer's own words, quoted, including the ambiguous ones. "Stop calling me at work" and "stop calling me" are different sentences and the difference is the whole matter.
- What did the collector do next? The following ninety seconds are usually more probative than the request itself.
- What was recorded in the platform, by whom, and when? The gap between the call timestamp and the status timestamp is a metric in its own right.
- What happened afterwards? Was there another call, another letter — and if so, under which of the permitted circumstances.
Every one of those is answerable from a transcript anchored to timestamps plus your platform's own audit trail. None of them is reliably answerable from a sampled QA sheet, because the odds that the one call containing the request is among the four reviewed that month are poor — the arithmetic is in why two percent sampling fails.
Ambiguity is the normal case
The tidy example — "I am asking you to stop contacting me" — is not what floors actually encounter. What they encounter is:
- "Don't call me at work." A limitation on a place, not a general request, and a different branch entirely.
- "I already told you people I'm not paying." A reference to a previous call that may or may not exist in your record.
- "This isn't my account." A dispute in substance, phrased as a denial of identity.
- "Just send me something in writing." Could be a validation request, could be a stall, could be a request for a payment plan on paper.
- "My attorney is handling this." A different branch again, with its own consequences.
Two things follow. First, your script needs a defined response for each branch, and the response for an ambiguous one should be to clarify, on the call, in plain words — because the clarification is itself the evidence. Second, the QA rubric should surface these as moments to review, not as verdicts. Our own rubric reports whether a cease request or a dispute appeared to be raised, with the timestamp and the quote that triggered it, and stops there. It does not decide that a cease request was legally made, because that decision has consequences a model should not be making. What a supervisor then does with the flag is its own discipline — covered here.
What to check on your own floor this month
Four checks, none of which need new software:
Take ten disputes from your platform log and find the calls. Time how long it takes. If you cannot locate the audio for all ten inside an hour, retrieval is your weakest link, and it is the one that hurts most under an eighteen-month-old complaint.
Measure the gap between call time and status time. Pull the last month of cease and dispute flags with both timestamps. A same-shift median is healthy; a scatter of multi-day gaps means the step is being done from memory at the end of the day.
Run the reverse test. Search a month of transcripts for the phrases — "stop calling", "don't call", "not my debt", "I dispute", their Spanish equivalents if your floor uses them — and check each hit against the platform. The calls that turn up here and not there are the whole point of the exercise. Without transcripts this test is not available to you, which is a fair thing to weigh in a buying decision.
Read your script's branches out loud. If a collector cannot say the cease-request response from memory, they will not produce it under pressure at minute nine of a difficult call.
The constraints, stated up front
Our capture is honest about what it does not do, and two limits bear directly on this topic.
Calls tie to collectors and times, not to accounts. We record who was on which call when; we do not know which account was open in your platform unless there is an integration. So we can show you the moment a dispute was voiced. Reconciling that against the right account is a join you or an integration performs, not something we can claim.
The call has to run through the computer. Audio comes from the collector's PC, which is what makes it work with any softphone. A desk handset that never touches the machine produces nothing to capture. If your floor is on physical phones, none of the transcript-side checks above are available to you and you should know that before you buy anything. Both constraints and the rest are published on compliance rather than left for you to find, and what actually gets captured is described on features.
As of September 2026
This describes the position as of September 2026. Statutory text and Regulation F are public at ecfr.gov and congress.gov; rulemaking history is at the Federal Register; the Consumer Financial Protection Bureau publishes the implementing materials and the Federal Trade Commission publishes enforcement and annual complaint reporting. State requirements and creditor contracts commonly add to the federal baseline and vary considerably.
Nothing here decides your obligations. It describes a records gap and how to test for it. If you want to know whether the reverse test would find anything on your floor, tell us how you dial and we will be straight with you about whether we can run it.
Frequently asked questions
Does a spoken cease request count?
That is a legal question with a federal baseline, state overlays and often creditor contract terms on top, and it is genuinely fact-specific — which is why we point at counsel rather than answering it. What is not in question is the operational half: if it was said on a call and you cannot retrieve that moment later, you cannot show what happened either way.
Why can't our dispute log answer this?
Because it records the outcome of a human noticing, not the noticing itself. A log with no missed disputes looks identical whether the floor catches everything or catches half. The calls are the only place the difference is visible.
What should a collector do with an ambiguous statement?
Clarify on the call, in plain words, and let the clarification be the evidence. "Don't call me at work" and "don't call me" are different branches, and guessing which one a consumer meant is the failure mode worth designing out of your script.
Does the software decide that a cease request was made?
No. It surfaces that a cease request or dispute appears to have been raised, with the timestamp and the quote that triggered it, and hands that to a person. Deciding what it was, and what follows, stays with people who can be held responsible for the decision.
How would we test our own gap without buying anything?
Take ten disputes from your platform log and time how long it takes to find the calls, then compare call timestamps against status timestamps for a month. Both are free. The third test — searching transcripts for the phrases and finding calls the log never got — needs transcripts.
We use desk phones, not softphones. Where does that leave us?
The voice side of this does not apply to you: audio is captured from the collector's computer, so a handset that never routes through the machine produces nothing. Desktop activity still works. It is worth settling which you use before evaluating anything in this category, ours included.
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.