Nobody objects to automated follow-up. They object to obviously automated follow-up, which is a different thing and entirely avoidable.
The tell is rarely that a message was sent by software. It is that the message does not make sense as something a person would send at that moment.
What to trigger on
The single biggest upgrade available is moving from time-based to behaviour-based triggers.
A time-based sequence sends message two on day three, to everyone, regardless of what happened. A behaviour-based sequence sends it because someone opened the first message and did not reply, or clicked the pricing link, or booked and then cancelled.
The second reads as attention. The first reads as a mailing list, because that is what it is.
Useful triggers, roughly in order of signal strength:
- Booked, then did not show. The highest-intent follow-up you will ever send.
- Clicked a specific high-intent link. Pricing, a demo, the signup page.
- Opened repeatedly without replying. Interested, hesitating.
- Went quiet after a conversation. The one most people never automate and where most revenue leaks.
- Nothing at all after capture. The fallback, and the weakest.
The shape of a sequence that works
Message one, immediately. Short, human, and useful on its own. Its job is to make sure nobody sits in silence. The response-time research is unambiguous about the cost of delay, and an immediate acknowledgement buys you the hours you need to reply properly.
Message two, a few days later. One specific useful thing. Not a "just checking in", which asks the reader to do the work of remembering who you are.
Message three. Address the actual objection. You know what it is; it is the same one every time in your market.
Message four, the honest close. "I will stop here unless you want me to pick this up." This performs far better than people expect, because it is the only message in the sequence that gives the reader an easy, dignified exit, and a surprising number take the other option instead.
Then stop. A sequence that never ends is a sequence people filter.
Writing so it does not sound automated
Short. Long automated emails read as templates because templates are where people put everything they might want to say.
One person to one person. No "we at". No "reaching out regarding".
Do not over-personalise. The uncanny effect comes from software demonstrating knowledge a human would not have volunteered. Referencing that someone opened an email three times is worse than saying nothing.
No fake urgency. Deadlines that are not real are the fastest way to become the sender people archive on sight.
Where the human has to show up
Automation should stop the moment a real signal arrives. Someone replies, books, or clicks something high-intent, and the sequence should end and hand over.
This is the single most common configuration failure. A lead replies "yes, interested", and two days later the software sends the scheduled "still thinking about it?" message. Everything the sequence built is spent in one line.
Whatever tool you use, check that reply detection actually removes people from the sequence, and test it with a real reply rather than trusting the setting. This is a place where having capture, pipeline and sending in one system helps concretely, because a reply logged against the contact is the same record the sequence is reading from. In a stitched-together stack that signal frequently lives in a different product.
The legal minimum, briefly
Automated commercial email carries obligations. In the US, the FTC's CAN-SPAM guidance requires accurate headers and subject lines, identification of the message as an advertisement where relevant, a physical postal address, and a working unsubscribe honoured promptly. Other jurisdictions have their own, frequently stricter, rules around consent.
Sequences that follow the advice above tend to be compliant almost incidentally, because short, honest and easy to leave is both better practice and better marketing.
Testing a sequence before it meets a real lead
Sequences fail in ways that are invisible from the editor, because the editor shows you messages rather than journeys.
Put yourself through it. Sign up with your own address, using a name you would notice, and let the whole sequence run. Read each message as it arrives, in the gap it actually arrives in, on a phone. Three things surface immediately that never surface in a preview: messages that assume a context the reader does not have, two messages arriving on the same day because two triggers fired, and the gap that is far too long.
Then do the more important test. Reply to message one, as a real lead would, and confirm that nothing else arrives. If message two turns up two days later, the sequence is not wired to stop, and that single fault will cost you more than the copy quality of every message in it.
The follow-up nobody writes
There is one sequence most affiliate businesses do not have, and it is usually the highest-value one available: the follow-up for people who went quiet after a real conversation.
These are not cold leads. They spoke to you, expressed interest, and then stopped, usually for reasons that have nothing to do with you. They are the warmest people in the pipeline and the least likely to be contacted again, because reaching out feels like nagging.
A two-message sequence, triggered by fourteen days of silence after a conversation, converts better than almost anything sent to a fresh lead. It works because it is genuinely reasonable: you spoke, it went quiet, and a short note asking whether it is still relevant is what a person would do.
Frequency, and knowing when to stop
The instinct is to keep going. Someone did not reply, so send another.
Two things argue against it. Deliverability degrades with engagement: mailbox providers watch whether people open and reply, and a list you keep mailing that never responds drags down the mail you send to people who do. And the marginal message performs worse each time, so message seven is not seven times better than message one, it is a rounding error attached to a real cost.
Litmus's long-running work on email return keeps landing in the same place: the return comes from relevance and timing, not from frequency. A four-message sequence sent to the right trigger outperforms a twelve-message sequence sent on a schedule, and it costs less reputation to run.
Set an end. When someone reaches it without engaging, move them out of active follow-up rather than into another sequence.
A sequence you can copy today
Trigger: a new lead arrives and has not replied.
- Immediately. Two lines. Acknowledge, say when you will follow up properly, and give them a way to jump the queue if they want to.
- Day two, if unopened. Same message, different subject line. Not a new message; people miss things.
- Day four, if opened but no reply. One specific useful thing related to what they asked about. No ask.
- Day eight. The objection you hear most, addressed directly.
- Day fourteen. "I will leave this here unless you want me to pick it up."
Every step stops the moment they reply or book. Five messages, two weeks, and it outperforms most of what gets sold as sophisticated nurture.
Common questions
How many follow-up messages should a sequence have?
Three to five for most affiliate offers. The first is an acknowledgement, the middle ones carry something useful, and the last one closes the loop honestly. Beyond five you are usually training people to ignore you.
Should follow-up be triggered by time or by behaviour?
Behaviour where you can, time as the fallback. A message that arrives because someone opened the pricing link is relevant; the same message on a fixed day is a broadcast that happened to land.
How do I stop automated emails sounding automated?
Write them as one person to one person, keep them short, and never reference something the software knows but a human would not have noticed. The tell is usually over-personalisation, not under.
When should a human take over from the sequence?
The moment someone replies, clicks a high-intent link, or books time. A sequence that keeps sending after a real signal is the fastest way to undo the goodwill it built.