The first three emails you can ship before you buy any tooling

Say you have sixty-odd signups. Maybe two hundred. Roughly a third of them did nothing after the first session. Meanwhile a browser tab sits open on somebody's pricing page comparing a $0 tier against a $45 one, while you try to work out which platform to commit to. That tab is the wrong tab. Close it for a fortnight.

The constraint at this stage is not sending capability. Your mail client can already send an email to one person, which is exactly the volume you have. The constraint is that you do not yet know what those emails should say, and no amount of tooling solves that. An ESP — an email service provider, the platform that stores your list and sends campaigns — is a delivery mechanism for messages that already work. Buying one first gets you a beautifully automated version of a message nobody replies to.

Here is what you are actually buying with the manual weeks: every reply is a piece of research you cannot get any other way. When someone answers your day-three nudge with "I got stuck on the import and gave up", you have just learned more about your activation problem than a dashboard would have told you in a quarter. Automate first and that reply never happens, because nobody replies to a no-reply address with a hero image in it.

Send it within one working day of somebody signing up. Not instantly. Instantly reads as automated even when it is not, and the whole value here is that it is not.

The shape: a subject line that names the thing they just did, an opening line that is not "thanks for signing up", one sentence of orientation and exactly one question. Something like:

Subject: you signed up for [product] yesterday
Hi [name] — I saw you created an account yesterday. I'm [your name], I built this. Most people come to us because [the one specific problem]. Is that what brought you here, or something else?

The question is doing the work. It is open, it is short and it is answerable in one line from a phone. "Let me know if you have any questions" is not a question. It is a closing formality, converting at roughly the rate you would expect from one. Ask something a person would answer at a bus stop.

Leave out the feature tour, the link to the docs, the calendar link, the discount. Every additional element you add divides the attention available for the one thing you actually want, which is a reply telling you why they showed up. The welcome sequence guide covers the fuller five-message version you will build later, once you know what belongs in it.

Target anyone who signed up, started setting up and stalled. Three days is a reasonable default; if your product has an obvious first-run flow, send it the day after they abandon it.

This email lives or dies on one property. It must name the specific step they did not finish. "Come back and finish setting up" is a nag. "You added your team but never connected the calendar — that's the bit that makes the reminders work" is a reason. You can see which step they stopped on by looking at your own admin view, which at two hundred users takes about eight seconds per person.

The structure that works, in three lines: the observed step, the one-line consequence of leaving it undone and the shortest possible path to doing it. Then stop. A second call to action in this email is you hedging, and the reader can tell.

Subject: the calendar bit takes about two minutes
You got as far as adding your team last week, then stopped before connecting a calendar. That's the step that turns the whole thing on — without it we can't see availability, so nothing schedules. Two minutes here: [link]. If it's the permissions screen that put you off, tell me and I'll do it with you on a call.

The last sentence is the one most founders cut. It is also the one that gets replies. Offering to do the hard part with them is only affordable while you are small, which is exactly why you should be doing it while you are small. Orbit's Copy Framework skill has the longer treatment of this structure — the observed behaviour, the consequence, the single next action — and it holds up when you eventually automate the message.

Goes at three or four weeks of no activity, to people who did use the product and then stopped. Not to people who never started. Those two groups look identical in a spreadsheet and need completely different messages: one has an unanswered question about whether the product works, the other has an answer you do not want to hear.

This email should be the shortest of the three and should not try to sell anything. You are asking a question and accepting the answer. A subject line as blunt as "did we drop the ball?" works here.

You were using [product] pretty steadily through August and then stopped. That usually means one of two things — it didn't do the job, or something changed on your end. Which one was it? Genuinely useful either way, and no follow-up if you'd rather leave it.

"No follow-up if you'd rather leave it" is not politeness. It raises reply rate, because it removes the suspicion that answering starts a sequence. Honour it — if you follow up anyway, you have taught that person that your emails are not to be trusted. You will never get the second answer.

Some of the replies here will be genuinely painful to read. That is the point. A churned user who tells you why is worth more than the four who left without a word, and at this size you can still email every one of them personally. That window closes and does not reopen.

Twenty minutes writing the three. Write them badly first and cut afterwards — the second draft of each should be shorter than the first, and if it is not, you edited rather than cut.

Fifteen minutes making them repeatable. Gmail calls these templates, Outlook calls them Quick Parts, Superhuman calls them snippets. Save each of the three with the personalised parts left as obvious blanks — square brackets are fine, because you are going to fill them in every time and a bracket you forgot is a bracket you can see. Do not build anything cleverer than this. Cleverness here is a tooling decision in disguise.

Twenty-five minutes clearing the backlog. Open your users list, sort by signup date and work backwards. Anyone from the last two weeks gets email one. Stalled accounts get email two. A month dormant gets email three. You will get through more than you expect, because each one is four sentences and thirty seconds of looking at their account.

Then put fifteen minutes in your calendar for tomorrow morning and do it again with yesterday's signups. The daily repetition is the actual program. Everything else in this post is preparation for the habit.

There is no analytics setup here and none is needed. Three columns in a spreadsheet: who you emailed, which of the three you sent and whether they replied. That is the entire measurement layer. It is more honest than the dashboard you would get from a platform.

Open rate is not available to you without tracking pixels. That absence is a gift. Apple's Mail Privacy Protection has been pre-fetching images on behalf of a large share of recipients since 2021, which means a meaningful chunk of your opens would be a machine, not a person. Skipping the metric entirely puts you ahead of teams still reporting it in board decks.

After two weeks the reply column tells you which of the three emails is worth automating first (the one with replies), which subject lines get answered and the two or three sentences people keep writing back that should become product copy. That is the brief for everything you build next.

If a message gets no replies across twenty sends, the message is wrong, not the channel. Rewrite the question. The most common fault is that it was not a question at all.

It has an end. Three signals, and you will usually hit one before you hit a signup count that matters:

Perhaps you missed the window more than once in a week — signups sitting three days before you got to them, which means the nudge arrived after the person had already forgotten you. Or you sent the same message verbatim three times running without changing a word, which means you have stopped learning from it. Or the trigger you need is something you cannot see from an admin list, like a behaviour deep inside the product.

Any of those and it is time. The line where automation starts paying covers what to do about it, and the short version is that you automate the one message you already send by hand and have started to dread — never a message you have not sent yet.

Before then, the order of what to build after these three is settled and worth reading once so you do not waste the first automated month on the wrong flow. That is the lifecycle for startups guide: welcome, then your conversion moment, then winback, in that order, for reasons that are about compounding rather than preference.

The one thing to do today: open your signups list, take the twenty most recent and send email one. Not a version of it you have polished for a week. The one you write in the next twenty minutes.