How to write a breakup email (the highest-replying touch)

A breakup email is the message that announces you are stopping. It is worth writing carefully, and it is worth being honest that the reputation attached to it — the highest-replying touch in the sequence — has never been measured in any source you can check. What is defensible is the mechanism: a message that removes the ask, names a specific condition for revisiting, and then is actually followed by silence. This is the structure, the arithmetic that explains the flattering statistic, and what has to stop in your system when it sends.

The claim in the name has no public evidence behind it

Start here, because everything else in this post is more useful once the headline claim is properly sized.

There is no authoritative public dataset of per-touch reply rates for cold email. No regulator publishes one, no standards body collects one, and no peer-reviewed study measures it. Every number in circulation traces back to a sending platform reporting on its own customers, with an undisclosed sampling frame and an undisclosed definition of what counts as a reply. Two vendors reporting on nominally the same metric routinely disagree by an order of magnitude, which tells you they are measuring different populations rather than converging on a truth.

There is also a mechanical reason the breakup's rate looks good that has nothing to do with the copy. Consider what the denominator is by the time you send it.

Take a starting list of 1,000 verified addresses. The numbers that follow are illustrative assumptions chosen to show the shape of the arithmetic, not benchmarks — substitute your own. Suppose the first touch produces some replies and those enrolments stop. Suppose bounces remove a slice, suppression removes another, and a handful unsubscribe. By touch six the population still receiving mail might be 850 rather than 1,000, and — this is the part that matters — it has been filtered by five prior opportunities to reply. Everyone easily persuaded has already gone. Everyone unreachable has already bounced out.

So the breakup's reply rate is computed against a smaller, harder, more-filtered audience, and any lift it shows is competing with a survivorship effect running in the opposite direction. That does not make the touch worthless. It makes the comparison across touch positions much less informative than it looks, and it means the only number worth quoting is your own, measured by position, on a fixed cohort.

What a breakup email actually is in a sequence

Strip the psychology and it is a terminal step: the last message before the enrolment ends, whose content acknowledges that it is the last one.

That definition is load-bearing because it tells you what the message must do. It is not a last attempt at persuasion dressed as politeness. If the copy is a final pitch with a valedictory sentence bolted on, the recipient reads a sixth pitch and nothing changes. The whole effect, such as it is, depends on the ask genuinely being withdrawn.

Three properties follow, and each one is a design constraint rather than a stylistic preference:

• It has to be credible. A message announcing an ending, followed by more messages, is a message the recipient now knows was untrue. The next sequence from your domain starts from a worse position than if you had simply stopped without saying so.
• It has to be short. Length signals continued investment, which contradicts the message. Two to five sentences is the working range, and the shorter end usually reads better.
• It has to leave a specific door open. Not "let me know if anything changes", which is a sentence with no content. A named condition — a renewal date, a hiring signal, a project the recipient mentioned — is the difference between a door and a wall with a door painted on it.

The mechanism people call loss aversion, sourced properly

The usual explanation is loss aversion, and it deserves both a real citation and a real caveat.

Loss aversion comes from prospect theory, set out by Daniel Kahneman and Amos Tversky in "Prospect Theory: An Analysis of Decision under Risk", published in Econometrica, volume 47, issue 2, in March 1979. The relevant finding is that the value function defined on gains and losses relative to a reference point is "generally steeper for losses than for gains" — a loss of a given size registers as larger than a gain of the same size.

Now the caveat, because it is the honest part. That paper studied choices between risky monetary prospects presented to experimental subjects. It did not study email, sales, attention, or a stranger deciding whether to reply to a message from someone they have never met. Applying it to a breakup email is an analogy, and a plausible one, but an analogy is a hypothesis about why something might work rather than evidence that it does.

Two adjacent mechanisms are at least as plausible and get discussed less. The first is pressure removal: several of the reasons a busy person does not reply — not wanting to negotiate, not wanting to say no, not having a decision yet — are cheaper to overcome once the reply no longer commits them to anything. The second is deadline clarity: an open-ended sequence has no moment that demands a decision, while a message that says "this is the last one" creates one. None of these is measured, and the practical implication is the same in every case, which is why the argument is worth less than the copy discipline it produces.

Six lines, and why each one is there

A breakup email is a small object. Every line has a job, and lines without one make it worse.

1. The withdrawal. State that you are stopping, in your own words, in one clause. "I will close the loop here" works. So does "I will stop chasing this."
2. The reminder of what it was. One line naming the specific thing you were offering. Without it, a recipient who skimmed the earlier touches has no idea what they are being asked to remember, and the door you are opening leads nowhere.
3. The named condition. The circumstance under which this becomes worth their time. Specific enough to be recognisable when it happens.
4. The re-entry instruction. How to restart it, which for a plain-text sales email is always "reply". Not a link, not a calendar, not a form. The point of the message is that it costs nothing.
5. No count and no guilt. "I have emailed you four times" converts your persistence problem into their obligation, and it is the single most common failure in this genre. So is "I will assume you are not interested", which is passive aggression in a tie.
6. A normal sign-off. Same one you use everywhere else. A special valediction for the breakup message reads as theatre.

What is deliberately absent is a call to action. Everything else in a sequence ends with an ask, and this one ends without one — that is the whole distinction, and if you find yourself adding "or just book 15 minutes here", you have written touch six rather than the breakup. The friction ladder that governs every other CTA in a sequence is the right reference for those touches; this one sits below the bottom rung on purpose.

Five drafts, and what changes between them

Templates are less useful than watching the same message adapt, so here is one situation written five ways. The situation: you sell an inventory-reconciliation tool, you have written five times, and the recipient runs operations at a mid-market retailer.

The plain version. "I will stop here so I am not adding to your inbox. If reconciling stock counts across your warehouses ever moves up the list, reply and I will pick it back up."

The version with a real condition. "Closing this out. The trigger that usually makes this worth doing is opening a second distribution centre — if that lands in your plan for next year, reply and I will send the thirty-minute version."

The version that gives something away. "I will leave it here. One thing you can use regardless: the reconciliation failures we see most often are unit-of-measure mismatches between the WMS and the ERP, and that is a config fix rather than a purchase. Reply if you want the specifics."

The version for a champion who went quiet. "I suspect this stopped being your priority, which is fine — I would rather stop than keep nudging. If it comes back, or if it has moved to someone else on your team, reply and point me at them."

The version that names the wrong-fit possibility. "Closing the loop. It is entirely possible this is not a fit — most of the teams this helps are running more than four locations, and if you are not, ignoring me was the right call. If that changes, reply."

Look at what varies. The withdrawal sentence is nearly constant. The condition changes with what you know about the account, and it is the line that carries all of the differentiation. That is the practical lesson: a breakup email is not a template problem, because the only line that matters is the one a template cannot write for you.

The fifth draft deserves a note. Naming the qualification threshold out loud costs you nothing with people who fail it and earns credibility with people who pass it, and it occasionally produces the most useful reply in the whole sequence, which is somebody telling you that you have the segment wrong.

What must actually stop when the breakup sends

This is the systems half, and it is where breakup emails most often break their own promise.

The enrolment ending has two distinct shapes, and they mean different things. An enrolment that runs off the end of its steps finishes as completed. An enrolment that ends early for a named reason finishes as stopped, carrying a reason string — a human reply writes replied, a hard bounce writes bounced, and an explicit terminal step writes sequence_end. Autocloz supports a stop step kind precisely so that the ending is a designed event rather than the accidental consequence of running out of steps, which matters when you later want to know why a lead is no longer being contacted.

Four things people assume happen at that point, and what actually does:

Ending a sequence does not suppress the address. Suppression is a separate row, per organisation, keyed on the address with a scope of either the exact email or the whole domain, and carrying its own reason. Only an unsubscribe, a bounce or a manual add writes one. A completed enrolment leaves the address perfectly eligible for the next campaign, which is correct behaviour and also the reason a breakup email means nothing unless list hygiene is handled separately.

A reply on another channel may not stop the email. In Autocloz the enrolment state machine is driven from email inbound: a human email reply stops the enrolment, a hard bounce stops it, and an out-of-office auto-responder deliberately leaves it running because the recipient is temporarily unavailable rather than disengaged. An inbound SMS does not write a stop reason on the enrolment. If you are running a multichannel cadence, know which inbound events your system treats as an ending and which it does not.

Suppression is keyed to the identifier, not to the person. A STOP keyword on a phone number writes a suppression against that number and stops SMS, WhatsApp and calls to it. It does not touch the email address or the LinkedIn profile belonging to the same human, because those are different identifiers. Suppressing a person everywhere means suppressing each identifier they reach you on.

The campaign does not close until every enrolment in it has ended. Which is worth knowing if you use campaign status as your signal that a cohort is finished.

The practical build is a terminal step whose content is the breakup and whose effect is a designed ending, with the re-approach handled as a new campaign later rather than as a resumed one. Composing that ending as an explicit last step is a five-minute job, and the six distinct states that "no response" can actually mean is the right diagnostic to run before you decide the breakup is the correct next touch at all.

Autocloz's free plan covers 5 users and 10 mailboxes with the sequence builder, suppression and reply handling included — start free and put a real terminal step at the end of your cadence.

The fake breakup, and why it costs more than it earns

The pattern is common enough to name. Send the breakup. Wait three weeks. Start a new sequence to the same person, with a new subject line, as though the first one had not ended.

The arithmetic looks fine in a spreadsheet. You keep the reply lift from the breakup and you keep the volume from continuing. The costs are real and land somewhere the spreadsheet does not measure.

It makes the next breakup inert. Whatever effect the message has depends entirely on the recipient believing it. A person who has received two of them from you has learned the format, and the third is read as a technique.

It converts an opt-out into a complaint. Someone who ignored five emails and then received a sixth after being told there would not be one has a specific reason to press the spam button rather than simply deleting. Complaint rate is the input that receiving systems weight most heavily, and this is a mechanism for raising it deliberately.

It contaminates your measurement. If a breakup is followed by a re-approach, you can no longer attribute the reply, and your own per-position numbers stop meaning anything.

The legitimate version of re-approaching is different in kind, not just in interval. Something has to have changed on their side — a role change, a funding event, a product launch, a hiring signal — and the new message has to open by referencing it. That is a new reason to write, not a resumed cadence, and working the dormant pile on a real trigger is a separate motion with its own rules.

What the law still requires in the last message

A breakup email is a commercial message, and nothing about announcing an ending exempts it from the rules that governed the previous five.

In the United States, the CAN-SPAM Act of 2003 requires accurate header information, a non-deceptive subject line, identification of the message as an advertisement, a valid physical postal address, a working opt-out mechanism, and that opt-outs are honoured within 10 business days. The Federal Trade Commission's compliance guidance states directly that the law makes no exception for business-to-business email.

Two practical notes for a final touch specifically.

The first is that "I am stopping" is not an opt-out mechanism. It is a statement of your intention, and it does not create a record, does not survive a change of sending tool, and does not bind a colleague who imports the same list next quarter. The compliant mechanism has to be a real one, and in Autocloz that means the unsubscribe footer plus the RFC 8058 one-click List-Unsubscribe header, which is on by default in the compliance settings, and the organisation's configured physical address injected at render time.

The second is jurisdictional. The rule that governs is the recipient's, not yours. The ePrivacy Directive 2002/58/EC treats direct marketing by electronic mail to natural persons as requiring prior consent while leaving other cases to national law, which is why the business-to-business answer genuinely differs across European Union member states. This is a summary for planning rather than legal advice, and your own facts deserve real advice.

How to measure whether yours is working

Three measurements, in ascending order of usefulness.

Reply rate by touch position, on a fixed cohort. Take one week of enrolments and follow only those. Count replies at each position against the number still active at that position. This is the only version of the headline claim that means anything, and it will differ from every published figure because your list and offer are not anyone else's.

Positive replies, not replies. A breakup reliably produces polite declines, and a decline is a genuine outcome — it frees the seat and updates the record — but it is not pipeline. Separate the two before you conclude anything, or you will optimise a copy variant that generates more courteous rejections.

The re-entry rate over ninety days. The named condition in the message is the thing being tested. How many recipients came back later and referenced it? This number will be small, and it is the one that tells you whether the door you left open was a real one. It also takes a quarter to read, which is why almost nobody measures it.

One trap worth flagging: do not A/B test the breakup against a sixth pitch and call the winner. They are different objects with different downstream effects, and a test that measures immediate replies while ignoring complaint rate and long-run domain reputation has measured the cheap half of the outcome.

What a breakup email cannot do, and what Autocloz does not do here

The limits, stated plainly.

A breakup email cannot rescue a sequence that was aimed at the wrong people. If nobody in the segment has the problem, the last message is polite rather than persuasive, and the fix is upstream in targeting.

It cannot recover deliverability. If the earlier touches landed in spam, so will this one, and the reply rate you measure will be a measurement of placement rather than of copy.

It cannot substitute for a decision about the list. Ending a sequence and removing someone from future outreach are different operations, and only the second one is a promise you can keep.

Autocloz specifically. There is no built-in "breakup step" as a distinct kind — you compose it as the final email step, optionally followed by an explicit stop step, and the honesty of the ending is a policy you enforce rather than one the software enforces for you. Nothing prevents you from enrolling the same lead in a new campaign the following week; the system will let you, and whether that is a genuine new reason or a fake breakup is a judgement it cannot make. Body-level A/B testing is not dispatched at all — variant testing applies to subject lines only, so a copy test between two breakup bodies is something you run by splitting the list yourself. And the reply classification that decides whether an inbound is a human reply, an out-of-office or a bounce is a model's judgement on a stranger's text; it is right often enough to be useful and wrong often enough that the queue is worth reading. Tools built purely as cold-email sequencers handle some of this differently, and the honest trade is set out on the Autocloz and lemlist comparison, while the email channel's own compliance surface covers the footer, the header and the address that every one of these messages carries.