Your email strategy should exist before your emails

Every week I pick a topic, break down a problem, and send it to you. This week I want to flip the direction. This issue is about you telling me what to write about.

Silent churn is costing you money. Merlin rewards your best readers with points, perks, and tiers. Apply for the beta at getmerlin.app

Most startup companies start off by writing the email and then worry about the logic later. They open the editor, write up something that seems okay, pick out a section that appears suitable, send it, and only then ask themselves why they sent the email to that individual at that time. In most cases when the outcome is a failure, the first thing people do is rewrite the copy; they blame the subject line, change the call to action, or suggest adding an emoji.

However, the various changes do not deal with the actual problem since the copy is not to blame; the true issue is concerned with the logic of the email.

You won't be able to produce an effective lifecycle email unless you are certain of the specific reason why that particular user should receive it at this moment rather than a different email or no email at all; writing the email should come last, not first. A great many teams take the opposite approach: they create the sequence first and then, if ever, consider the logic.

Before you compose the email, figure out the situation.

Before you write anything, make sure you know what’s really happening for each email you want to send:

It looks obvious when it's written out, but in reality it rarely takes place. I've seen a lot of sequences which include a welcome email, a series of nurture emails, and a 'we miss you' email, but in none of these cases were the emails linked to what the user actually did using the product. Rather, they were sent according to a schedule - on day 1, day 3, and day 7 and whether or not the user received the email had nothing to do with their activities.

Why this matters more for AI products

With most SaaS products signing up generally indicates that a person is interested, things are different with AI products. It's possible for a user to set up an account, make one generation, reach a usage limit, invite a colleague, upgrade, cancel and then come back all within just a few days. Signing up and actually activating the product are two distinct actions, and it is precisely because of this confusion that so many emails end up being wasted.

I have witnessed this occurring personally: with a free-tier product, a large number of the people who signed up never produced a single result. A general "welcome, here's what you can do" email wasn't effective because the real problem wasn't a lack of awareness it was that these users had never used the key feature. The solution was to divide the audience according to their level of usage: those who had never generated anything received one message, those who had generated once and then stopped got a different one, and the active users were not sent anything since they didn't need a reminder: the same list, but three different scenarios, and three different responses.

I have also come across another kind of problem: during the launch the drop-off had no connection with email opens or clicks. Clicks were acceptable and so were the number of signups. The actual failure took place at checkout, since people started the process but then left before making an attempt to pay. It will not be of any help to make changes to the email logic here, as the problem is entirely outside the email stage. That is one of the reasons why this exercise is important: when you establish the real triggers you also quickly realize which problems email cannot deal with, and thus stop wasting time rewriting subject lines in an attempt to fix a bug at checkout.

What this looks like in practice

Once you figure out the real situations, your “sequence” becomes a short list of actual triggers:

The various situations differ and each has its own reason for sending or not sending an email. You won't need to produce more creative copy; you just have to work out which situation applies before you begin writing.

The logic comes first.

That is all that people notice in the email; the real work the aspect which enables your lifecycle program to be effective occurs before this stage. You need to decide what should be considered a genuine trigger, which individuals should be excluded, and what happens next according to what the individual Before you write 20 emails, figure out the 20 different situations in which a user actually needs one.ds one.

This is the first entry in a series in which I examine the errors that AI startups make both before and after their launch; in the following article I will explain why your domain architecture forms part of your product infrastructure and not just a matter of deliverability.

Before You Go: Here's How I Can Help

— Pavel