How to Build an Email Marketing Strategy From Scratch
The six decisions that have to be made before the first send, in the order that stops you rewriting the first three.
On this page
- Decision one: who this is actually for
- Decision two: the promise
- What a promise looks like written down
- Decision three: cadence you can hold in a bad month
- Decision four: what subscribing leads to
- Decision five: the number that decides continuation
- Decision six: the stack, last
- Before your first send
- Frequently asked questions
The short version
- Six decisions have to be settled before the first send, and they constrain each other in one direction only — audience decides promise, promise decides cadence, cadence decides everything downstream.
- The most common failure is starting at the tool. A provider chosen before the promise is written gets picked on price and feature lists rather than on what the programme actually needs.
- Write the promise as a sentence a subscriber would recognise. If you cannot, the sequence has no shape and every send becomes a separate decision.
An email programme built without these decisions does not fail loudly. It drifts. Six months in there is a newsletter going out most Thursdays, an open rate somewhere in the twenties, and no one able to say what would count as it working. Nothing is broken enough to fix, which is exactly why it never gets fixed.
The six decisions below are ordered because they constrain each other in one direction. Audience narrows the promise. The promise sets the cadence you can sustain. Cadence determines how much you need to produce, which sets the budget, which limits the stack. Work them out of order and you will redo the first three.
Decision one: who this is actually for
"Our customers" is not an audience. It is a database. An audience is a person with a specific situation, and the test is whether you can name something they know that a general reader does not, and something they are worried about that a general reader is not.
This matters immediately because it decides your baseline. Writing for people who already run email programmes means you can use the word suppression without explaining it. Writing for people about to start one means half your first sequence is vocabulary. Both are viable. Trying to serve both in one list produces email that is too basic for one group and too vague for the other.
If your list genuinely contains both, that is a segmentation problem rather than a strategy problem, and the fix is a question at signup rather than a compromise in the writing.
Decision two: the promise
The promise is a single sentence stating what arrives, how often, and what it does for the reader. It is the thing the opt-in form says, the thing the welcome email confirms, and the standard every subsequent send is measured against.
The reason to write it down is that it makes rejection possible. Without a promise, every idea is arguably on-topic, so nothing gets cut and the programme becomes whatever was easiest to produce that week. With one, an idea either serves it or does not.
What a promise looks like written down
The difference is not tone. It is whether the sentence can reject anything.
Too loose to use
Nothing fails this test
- "Tips, news and updates from the team."
- "Insights to help you grow."
- "Our latest thinking on marketing."
Specific enough to enforce
Most ideas fail this test, which is the point
- "One deliverability problem, diagnosed and fixed, every second Tuesday."
- "A teardown of one real opt-in form each Thursday, with the numbers."
- "Monthly: what changed at the mailbox providers and what you have to do about it."
Decision three: cadence you can hold in a bad month
Choose the frequency you can sustain during your worst month of the year, not your most enthusiastic one. A weekly promise kept for six weeks and then abandoned trains subscribers to stop expecting you, and re-establishing that expectation costs more than never having set it.
Frequency also interacts with deliverability in a way that catches people out. Long gaps between sends are a reputation risk, because a list that has not been mailed in three months contains addresses that have gone stale and engagement patterns the mailbox providers no longer have data on. Sending monthly is fine. Sending monthly, then nothing for a quarter, then a burst is not.
Decision four: what subscribing leads to
Every list has an offer path even when nobody wrote one down. The question is whether it was designed or whether it emerged. Designing it means deciding what you eventually want a subscriber to do, how long that typically takes, and therefore how many sends happen before the first ask.
The number that matters here is time to first purchase, not conversion rate. A programme where people buy in three weeks and one where they buy in eleven months need different sequences, different content and different patience from whoever is funding it.
Decision five: the number that decides continuation
Pick one metric that would justify keeping the programme running, and pick it now rather than at the first review. Deciding afterwards means picking whichever number happened to look good, which is how programmes survive for years without producing anything.
For most lists this is revenue attributable to email, or qualified enquiries, or subscriber-to-customer rate. It is very rarely open rate, which measures whether people recognise your name rather than whether the programme works.
Decision six: the stack, last
By this point the tool question mostly answers itself. You know your list size, your cadence, whether you need branching automation or a single linear sequence, and whether anyone on the team can write HTML. That is enough to rule out most options without reading a comparison table.
Choosing the tool first inverts this. You end up designing the programme around what the software makes easy, which is how organisations end up with elaborate automation nobody asked for and no clear promise.
Before your first send
- The promise is written as one sentence and appears on the opt-in form
- The cadence is one you held through last year's busiest month
- SPF, DKIM and DMARC records are published and verified
- A welcome email exists and confirms the promise in the same words
- One success metric is written down and agreed with whoever funds this
- There is a plan for what happens to subscribers who never open
Frequently asked questions
How big does a list need to be before this is worth doing?
Small lists benefit more, not less. With 300 subscribers a bad promise costs you almost nothing to fix; with 30,000 it means rewriting an expectation you have spent two years setting.
Can the promise change later?
Yes, and it should when the audience shifts. Announce it in a send rather than letting it drift, because an unannounced change reads to subscribers as the programme getting worse.
What if the offer path is genuinely long, like enterprise sales?
Then measure something that moves on a shorter horizon — reply rate, meeting requests, or progression between segments — and treat revenue as the annual number rather than the operating one.