The follow-up cadence that gets replies

In this guide

1. 01How long should you wait between follow-ups?
2. 02How many touches before you stop?
3. 03How do you vary the message so it does not read as automated?
4. 04When do you stop — and what actually counts as being told to stop?
5. 05Does the cadence change for past clients and your sphere?
6. 06How I actually run it

How long should you wait between follow-ups?

Four to six weeks between touches on a cold list. That is the answer, and the rest of this section is why. If you are texting a skip-traced list every three days and getting silence back, the spacing is doing more damage than the script is. This post is about touches two through six — the first one is its own problem, and it is covered separately in the first text to a motivated seller.

Before I built software I ran a restoration company. It went from zero to a little past $500k in its first twelve months with no ad budget, on roughly $3,600 a year of text messages. The cadence was every four to six weeks, to everyone who had ever given me a number, indefinitely. Two honest notes about that, because it is the part people quote back at me. It was not clever — it was just the only part nobody else kept doing. And it was a list of people who had handed me their number themselves, run with no end date, which is a different posture in both directions from a skip-traced list on a fixed sequence. The interval is the part that transfers. The rest of this post is why the other two do not.

There are two constraints, and the interval is where they meet. Under about two weeks, your last message is still sitting unanswered in the thread — a second one on top of it does not read as persistent, it reads as pressure, and it is the fastest way to convert a maybe into a block. Past about eight weeks, the thread has fallen off the screen and you are introducing yourself again from zero, which wastes the only asset a follow-up sequence has: the fact that they have heard from you before.

You will see a statistic quoted at you constantly — that some large majority of sales happen after the fifth follow-up. I have gone looking for the study behind it more than once and the trail always runs cold, so I am not going to build your cadence on it. What I will tell you is the part I can stand behind: the fourth message has produced more real conversations for me than the first one ever did, and the only reason it works is that almost nobody is still there at four.

Once they reply, the cadence is over

Everything above is a schedule for silence. The moment somebody answers, the schedule is finished and you are in a conversation, which runs on conversation time — minutes and hours, not weeks. The most expensive mistake in a follow-up system is letting a live reply sit in a queue because the schedule said the next touch was five weeks out. Answer it, then take that person out of the sequence entirely so no scheduled message ever lands on top of a real exchange.

How many touches before you stop?

Five to seven on a cold list — and you pick the number before the first message goes out, not when you get bored. Deciding in advance is what turns a pile of texts into a sequence you can actually judge. If a list is on touch six and has produced nothing, that is information. If the list has been running forever at no particular interval, you have learned nothing and you cannot stop, because there is no line to stop at.

The persuasion argument for a cap is simple. Somewhere past the fifth or sixth unanswered message you have stopped changing anybody's mind and started manufacturing complaints. Nobody has ever sold a house because the eleventh text wore them down.

The rule that makes cadence a legal question

This is the part a cadence post usually skips, and it is the reason the cap is not only a taste question. The federal rule that reaches cold outreach is not the autodialer ban — it is the Do Not Call registry, which has no equipment element at all, so how you sent the message is beside the point. And the private claim under 47 U.S.C. § 227(c)(5) is written for a person who received more than one call or message in a twelve-month period. Read that against a six-touch sequence. Touch one is not the claim. Touch two is what makes it one, and touches three through six are the multiplier. The discipline this post is built on and the exposure are the same behavior described from two directions.

The registry rule also comes with a safe harbor that is entirely paperwork, which is worth knowing precisely because it is conditional rather than automatic: written procedures, trained people, records, and a version of the registry obtained from the administrator no more than 31 days before the message goes out. Sellers and telemarketers subscribe through the FTC's telemarketer portal; it is not a file you download once and reuse for a year. Whether and how any of that applies to an offer to buy somebody's house is the whole subject of what TCPA actually says about texting a skip-traced list, and it is a live argument rather than a settled one. The point here is narrower: scrubbing and cadence are one decision, and only one half of it is a marketing choice.

The carrier side pushes the same direction for different reasons, because this audience works lists it did not collect. Silence is not consent — a number that has never replied has told you nothing. CTIA's Messaging Principles and Best Practices tell message senders not to use opt-in lists that have been rented, sold or shared, and to create and vet their own; carriers grade your traffic against that document. It does not make a purchased list unlawful by itself, and CTIA is not law. It does mean the burden of being conservative sits with you on both tracks at once, and the cheapest way to carry it is a sequence with a hard end.

When the sequence runs out, there are only three honest options:

1. Drop them. Off the active list, no further sends. The cleanest and most underrated option.
2. Move them to a long cycle. Two touches a year, not six weeks — and only for people who never asked you to stop and never bounced. This is a different list with a different cadence, not the same sequence running slower.
3. Move them to a channel they chose. If somebody has told you they would rather get a call or an email, that is a preference they expressed. Honor it and stop texting.

What is not on the list: starting the sequence over from touch one, six months later, as if the first six never happened. They happened. They are still in the thread on their phone, and the twelve-month window in § 227(c)(5) does not care which campaign you filed them under.

How do you vary the message so it does not read as automated?

Change the ask, not the adjectives. Most people vary a follow-up by rewording the same request — are you interested becomes are you still interested becomes just checking if you are interested. That is the same message three times and it reads exactly like what it is. A sequence feels human when each touch is doing a genuinely different job.

1. Vary the ask. Touch one asks for a yes. Touch three asks a question they can answer without committing to anything. Touch five asks about a number. Touch six asks for permission to go away.
2. Vary the length. A long message and a seven-word message feel like they came from different moments, because they did.
3. Vary the specificity. Early touches can be general. Later ones should name the street, the timeframe, the thing only somebody paying attention would mention.
4. Vary the angle. Condition, timing, taxes, an inherited property, a tenant who left. Each one is a different reason a person might finally answer.
5. Never vary the exit. Every message gets a plain way out, every time. This is the one element that should be boringly identical.

Three mechanical rules hold across all of them. Say who you are in every single message, not just the first — a name in the thread from six weeks ago is not identification today, and for a telephone solicitation 47 CFR 64.1200(d)(4) is written in terms of the individual caller's name, the name of the entity on whose behalf the message is sent, and a phone number or address where that entity can be reached. The sample sequence above carries the first two in every message; the third is the element people drop, and it is worth knowing the rule names it. Understand what identification buys you and what it does not: it satisfies an identification requirement. It does not create consent, and it does not get you past the registry. Keep one idea per text, and keep it near one segment (about 160 characters) where you can; it is cheaper and it reads like a person rather than a press release. And if you put a link in, put it on a domain that is yours — CTIA's guidance is that where a URL shortener is used, senders should use one whose web address and IP are dedicated to that sender exclusively, and public shorteners are among the most reliable ways to get an otherwise fine message filtered.

When do you stop — and what actually counts as being told to stop?

Immediately, and everywhere you send from. The part most people get wrong is not the timing — it is what they think counts as a stop request in the first place. (For what should happen mechanically the moment one lands, see what happens when someone replies STOP.)

One precision first, because "stops everything" is the kind of thing that gets stated harder than it can be. That a revocation on one kind of message also kills unrelated messages from you is part of the FCC's February 2024 revocation order, but that specific piece has been delayed more than once and is currently set to take effect January 31, 2027. So treat one-stop-stops-everything as the operating standard worth running and the one you can actually prove you ran — not as a rule you can date to today.

On what counts, the same order is specific. Replying to a text with stop, quit, end, revoke, opt out, cancel or unsubscribe is a per se reasonable means of revoking consent — those seven words are settled and take no argument. Everything else is not a bright line, and the rule does not say other wording fails. It says the question is decided on the totality of the circumstances, and a sender who wants to treat a reply as something other than a revocation carries the burden of showing that. Two companion pieces of the same order matter just as much day to day: a consumer may revoke by any reasonable method, and a sender may not designate an exclusive means of doing it. Which is why reply STOP to unsubscribe is a helpful instruction rather than a rule you get to enforce against somebody who wrote a sentence instead of a keyword.

So the interesting replies are the ones automation does not recognize. Some of these are revocations under the rule; others are not revocations at all and are still reasons there is nobody to text:

• take me off your list, lose my number, do not text me again — different words, unmistakable meaning, and arguing otherwise is a burden you would be carrying.
• not selling, not interested, we are not moving — read in context these are the same request, which is exactly the sort of judgment the totality-of-the-circumstances test is asking you to make honestly.
• wrong number — not a revocation, because the person you reached never gave consent to revoke. It is a mistraced or reassigned number, which is a separate federal problem with its own database. Either way there is no seller at that number.
• who is this followed by no further reply — not a stop in any legal sense, but it is a person telling you something.
• A phone call, a voicemail or an email saying the same thing. Nothing requires the revocation to arrive in the channel you prefer.

On timing, the rules set an outer bound rather than a target: honor a revocation within a reasonable time, not to exceed ten business days from receipt, and the same ten-business-day ceiling now applies to a company-specific do-not-call request. Ten days is not a budget. There is no operational reason a stop should take more than a moment, and the gap between the request and the next scheduled message is precisely where the complaints and the demand letters come from. The internal-list paragraph of the same rules is also where the record-keeping lives: an entity making telephone solicitations is expected to have written do-not-call procedures and trained people before the first send, to record a request at the time it is made, and to honor it for five years. Five years is longer than any spreadsheet survives, which is the practical reason the record belongs in one system instead of in your head.

Does the cadence change for past clients and your sphere?

Yes — longer between touches, and the ask gets smaller. Monthly to quarterly is the honest range for people who already know you, and the reason has nothing to do with the rules. A cold list tolerates a repeated ask because there is no relationship to damage. A past client does not: three asks in a row and you have converted a referral source into somebody who reads your name and feels sold to. The right frequency for a warm list is whatever cadence you can sustain while having something to say that is not a request.

One piece of law worth stating precisely, because it gets waved around loosely. The rules define an established business relationship as one formed by voluntary two-way communication, on the basis of a purchase or transaction within the eighteen months before the call, or an inquiry or application within the three months before it (47 CFR 64.1200(f)(5)). Three things follow, and the third is the one that gets missed. It expires — eighteen months after a closing, that particular basis is gone. It ends the moment that person asks your company to stop, even if they keep doing business with you otherwise. And it is an exemption from the definition of telephone solicitation, which means it does not touch the internal do-not-call obligations at all: the written procedures, the record and the identification apply regardless of how warm the relationship is. It is a narrow carve-out with a date on it, and it is one of the first things a lawyer will ask you to document — not a general permission slip that answers every consent question about every kind of message.

How I actually run it

The cadence is not the hard part. Remembering it in month four is the hard part, and it is the only part that decides whether any of this works. My whole system was a date, a list, and the discipline to send when the date came up whether or not I felt like it. Everybody can write a good first text. Almost nobody is still sending touch four eighteen weeks later, which is the entire reason touch four works.

Full disclosure: reitexter is mine, and it exists because I got tired of running that loop out of a spreadsheet. Since the rest of this post is about being precise, here is what it does. A campaign is scheduled for a date in the future, repeats on a cycle you choose, and ends after a set number of runs, so the sequence gets the hard stop this post keeps arguing for. Templates keep each touch a deliberate rewrite instead of a copy-paste. Opt-out state is re-checked at send time rather than only when the list was built, so somebody who says stop between touch three and touch four does not get touch four — and because that check reads the contact record rather than a snapshot of the list, re-importing the same CSV does not resurrect them either: the importer matches on the phone number and updates the existing contact instead of creating a second one. Every plan has every feature; the tiers differ by numbers, segments and seats. Pricing is here if you want the numbers.

And here is what it does not do, because a compliance post that oversells its own software is worth nothing:

• The schedule runs in one time zone — the campaign's, not each recipient's. Nothing in the send path knows where a given phone physically is. A list that spans zones is a campaign per zone, and that is a decision you make, not one the software makes for you.
• The automatic opt-out catches two words. An inbound reply is suppressed automatically when it is a single word and that word is stop or unsubscribe. Take me off your list arrives as an ordinary message and a person has to mark it — which is the whole reason the section above spends so long on wording rather than on keywords.
• The suppression gate sits on the automated paths. Campaigns, Bulk Send, the queued sender, the missed-call auto-text and the AI responder all put every recipient through one check — blocked, unsubscribed, a recorded STOP, or membership in the unsubscribers group — before anything leaves. A message you type by hand into a one-to-one thread does not go through that gate: the thread shows you the contact's unsubscribed and blocked state and then sends what you typed, because a human is reading the conversation. That is a deliberate choice, and it is also the seam to know about.
• It does not scrub the national Do Not Call registry, and it does not enforce calling hours. No software can tell you whether your funnel makes a given message a telephone solicitation, either. Those stay yours.

Whatever you run it in, the shape is the same: four to six weeks, a number of touches you chose in advance, a different message each time, and an instant stop the moment somebody asks. The people who do that outlast everybody who is better at writing the first text.

Frequently asked questions

Sources

• 47 U.S.C. § 227 — the TCPA, including the private do-not-call action at § 227(c)(5)
• 47 CFR § 64.1200 — calling hours, do-not-call lists, revocation, established business relationship
• FCC 24-24 — Report and Order on revoking consent for robocalls and robotexts
• CTIA — Messaging Principles and Best Practices (May 2023)