How to build an email list from scratch, and keep it

What building an email list actually means

An email list is a set of addresses whose owners have asked to hear from you, held somewhere you can send to them from. That is the whole definition, and two words in it do the work. Asked, because an address collected any other way is a liability rather than an asset. And held, because a list that exists only inside a platform you do not control is a list you can lose in an afternoon.

Building one is therefore not a growth activity in the way it is usually described. It is a trade. Somebody hands over a working address and the attention that comes with it; you hand over something worth that. The trade is the mechanism, and almost every list that never grows failed at the trade rather than at the traffic: nobody could tell what they were getting.

Which means the order of work is fixed. Decide what you are giving. Put the ask where the people who want it already are. Prove they asked. Deliver what you promised, immediately. Only then is there any point measuring, and only then does more traffic produce more subscribers rather than more bounces.

Starting this weekend: the seven steps, in order

Starting an email list takes seven steps and about a weekend. Do them in this order, because six of the seven are hard to change once addresses are arriving, and the sections below go through each one in more depth.

The order that works:

• Write the promise in one sentence, before anything else: what the reader gets, and by when.
• Set up a sending identity that authenticates, on a domain you control.
• Build one page whose only job is the offer, with a one-field form.
• Make every sign-up confirm before anything else is sent.
• Write the first message before you collect one address.
• Ask the people you already know, one at a time.
• Measure the four stages, then repeat the one that is weakest.

Why an address is worth more than a follower

A follower is a permission that somebody else controls. The platform decides how many of them see a given post, and that decision changes without notice and without appeal. An address is a permission the two of you hold between you: nothing sits in the middle deciding whether today is a day your reader hears from you.

That is the durable argument, and it survives every change in fashion. The fragile argument, the one about reach numbers, is not worth making, because reach numbers on a list are also worse than they look: a healthy list does not deliver to everybody, and an unhealthy one barely delivers at all. The point is not that email reaches more people. It is that the relationship is portable, addressable one person at a time, and does not evaporate when an algorithm is retuned.

The cost of that permission is real and worth naming up front. You take on a duty of care: a record of what each person agreed to, a way out that works on the first click, and an obligation not to send them things they did not sign up for. Those duties are the reason a list retains value. A channel anybody can shout into is worth nothing to shout into.

Decide what you are giving, before anything else

The single largest determinant of how many people join is what joining gets them, stated specifically enough to be evaluated. Subscribe for updates is not an offer; it asks the reader to guess. A named thing with a shape they can picture is an offer, and the difference between the two is usually several times the sign-up rate, not a few per cent.

A finite, delivered-by-email thing has an advantage that a download does not: it arrives in the inbox, so the first thing it teaches your reader is that mail from you is worth opening. A PDF teaches them that mail from you contains links to files. Both get the address. Only one of them gets the next open, which is the asset you were actually collecting.

Whatever you choose, write down the promise in one sentence before you build the form, and then make the form say it. If you cannot write the sentence, the offer is not ready, and no amount of work on the page will cover that. The lead magnet guide and the email course guide take this decision further.

A promise is specific enough to put on a form when it names all three of these:

• The thing itself, in the words the reader would use. Not a resource or a guide, but the five emails that get your first paying client, or the seven-day plan for a dog that pulls on the lead.
• How much of their life it costs. A number of days, a number of minutes a day, or a page count. An unbounded promise is a reason to postpone.
• What they will be able to do at the end that they cannot do now. If the answer is know more about a topic, keep going until it is a verb.

Put the ask where people already want it

A form converts according to why the person is looking at it, far more than according to how it is designed. A page whose only job is the offer converts several times better than the same form in a sidebar or under an article, and the gap is intent, not layout: one reader followed a link about the offer, the other came for something else. The planning band for each placement, and the traffic each one needs, are in the guide to getting more subscribers.

So the highest-value placement is a page of its own, linked from everywhere the offer is mentioned. Then the end of the piece of writing the offer extends, where a reader who has just finished something has a reason to want the next thing. Then, some distance behind, the sitewide placements: a footer, a header bar, a prompt on exit. Those are worth having and are not worth optimising.

One field, and what a second one costs

Ask for the email address. Nothing else, unless a later message is genuinely different because of the answer. A first name field is commonly reported to cost around a tenth of sign-ups, and on a sequence that reads identically for everybody it buys a greeting that most readers do not notice and some find slightly wrong.

The usual defence of a name field is personalisation, and it is worth testing against what personalisation actually produces: an address typed as a nickname, a job title, a blank, or the word test. A greeting built from that field is then wrong for a meaningful share of readers, which is a worse outcome than no greeting. If the name matters to you, ask in the first message and let people answer who want to.

Where a second field is genuinely load-bearing — a country that changes what you may lawfully send, a role that changes the content — put it on the form and accept the cost knowingly. The rule is not never ask; it is never ask for something nothing will use.

Label the field and the button with what they do

The button should name what arrives: send me the five emails, start the course, get the checklist. Subscribe describes what the reader is agreeing to become rather than what they are about to receive, and submit describes what the software is about to do. Both are weaker than a sentence about the thing.

The field needs a real label, not only a placeholder. A placeholder disappears the moment somebody types, which leaves a person who was interrupted mid-form looking at a box with no idea what belongs in it, and screen readers treat a placeholder inconsistently. A visible label above the field costs one line and removes a whole class of abandonment.

Keep the whole thing operable without a mouse and wide enough to press on a phone. Set the input type to email so a phone shows the right keyboard, do not disable paste, and do not make the field so narrow that a long address scrolls out of view while it is being checked.

The parts of a form, and what each one is actually for:

• A heading that repeats the promise in the same words as the page it sits on, so a reader who scrolled straight to the form still knows what they are getting.
• One labelled field, type email, with no placeholder standing in for the label.
• A button naming the thing that arrives, in the first person where it reads naturally.
• One line under the button saying what happens next, including that a confirmation is coming and the name it will arrive from.
• A link to how the address will be used and how to leave, close enough to the button that it is part of the decision.
• An error message next to the field, in words, saying what to change. Not a red border and nothing else.

Prove they asked, then keep the proof

Consent is the part that people skip because nothing appears to break. It breaks later, all at once: a complaint rate that puts every message you send into a spam folder, or a regulator asking a question you cannot answer. The rule that matters in every jurisdiction is the same one, and it is about evidence. You have to be able to say who agreed, to what, when, and by what action.

The strongest version is confirmed opt-in, which most people call double opt-in: nothing at all is delivered until the address owner clicks a link in a first message. It costs sign-ups, and the cost is known — expect a fifth to a third of sign-ups never to confirm, which is the range commonly cited in published double opt-in comparisons. What it buys is that every remaining address is provably a person who acted twice, and that a mistyped or malicious address never receives a second message.

In the European Union and the United Kingdom the GDPR requires consent to be freely given, specific, informed and unambiguous, and requires you to be able to demonstrate it. In the United States CAN-SPAM does not require opt-in at all, but does require an accurate From line, a working unsubscribe honoured promptly, and a physical postal address in the message. The large mailbox providers add their own conditions on top for anyone sending in bulk, and theirs are the ones that decide delivery day to day.

Whatever the jurisdiction, a consent record that answers a complaint holds all four:

• The address, and the exact wording the person agreed to, kept as it was worded on the day rather than as it reads now.
• When they agreed, as a timestamp, and when they confirmed if confirmation was used.
• How they agreed: which form, on which page, from which campaign. A record saying website is not a record.
• What has been sent to them since, and whether they are still subscribed, so a request to be forgotten can actually be carried out.

Where subscribers actually come from

Traffic is the part everybody starts with and the part that matters least until the three decisions above are settled. It also behaves differently from the way it is usually discussed: the channels that produce the most subscribers per hour of work are almost never the ones that produce the most visitors.

Read the table by the last column. A channel that sends people who were already looking for the thing you are offering converts in the tens of per cent. A channel that interrupts somebody converts in the ones. Both are useful, and mixing them up is how a month disappears.

One channel deserves separating out, because it is the one most often skipped: the people you already know. Past customers, people who replied to something, the contacts in your own sent folder. They cannot simply be added — that is not consent, and in the European Union it is not lawful — but they can be asked, individually, and they convert better than any advertising you will ever buy. Which channel to add next, once the page converts, is in the guide to getting more subscribers.

What has to arrive once somebody says yes

The moment after the form is submitted is the most attention this person will ever give you. Almost every list wastes it. The message that arrives is a receipt, or nothing arrives for a day, and by the time something does the reader has forgotten which of forty forms this was.

So deliver the promised thing first, immediately, in the first message (what to send new subscribers takes that message apart), and in the message body rather than behind a click wherever that is possible. Then keep going for a few days while the attention lasts: a short, finite sequence that ends is read far more completely than an open-ended one that does not, because a reader can see the end of it and has a reason to stay to it.

After that, the list needs a rhythm it can keep. The commonest failure here is not too many emails; it is an irregular silence broken by a sales message. A reader who hears from you every week and then gets an offer understands what happened. A reader who hears nothing for five months and then gets an offer marks it as spam, and on a shared sending reputation that damage is not only yours.

What to count, in the order that tells you something

Count the stages downward, not the total. A list size is a vanity number that hides which part is broken; the same list size can be a healthy funnel with thin traffic or heavy traffic through a broken form, and the two need opposite work.

Do not use open rates for anything that matters. Apple Mail Privacy Protection loads remote images through a proxy whether or not the reader looked, so on any consumer list the open rate rises when nobody reads anything. Define engagement as a click on a link a reader who actually did the thing had a reason to press.

The two numbers to watch for danger rather than for growth are the complaint rate and the bounce rate. Both are reported by the mailbox providers and both are conditions on your mail arriving at all, which makes them the only metrics on the list that can take the whole channel away; the deliverability guide sets out where to read them.

In order, and each one only worth reading once the one above it is healthy:

• Visitors to the page that carries the offer. If this is near zero, nothing below it is a real number.
• Submissions as a share of those visitors. Below the band for that placement type, the promise is not specific enough or the traffic was never interested.
• Confirmations as a share of submissions, if confirmation is used. A low figure means the confirmation message is landing badly or was never expected.
• Clicks on the first message you send. This measures whether the promise was delivered, and it is the earliest honest signal you get.
• Complaints and bounces, as a rate, watched against the thresholds the mailbox providers publish rather than against last month.
• Unsubscribes per send, read as a shape over time rather than as a number. A spike names the message that broke the promise.

The four ways list building destroys a list

Each of these is common, each looks like growth for a few weeks, and each ends the same way: mail that no longer arrives, for the sender and sometimes for everybody sharing a sending reputation with them.

Buying or renting a list is the fastest. The addresses did not ask for you, so complaints arrive immediately, and the spam traps seeded in every purchased file mark the sender directly. No sending platform worth using permits it, and finding out which one does is not a strategy.

Scraping addresses from directories and social profiles is the same act with more work. Adding customers because they bought something once is subtler and just as damaging: a transaction is not a subscription, and in the European Union the exemption for existing customers is narrow, specific to similar products, and still requires a way out offered at the point of sale. And importing an old list nobody has heard from in two years sends a message to addresses that have since become traps, which is why every import should ask each address again rather than assume, as the guide to importing a list describes.