Digital marketing

Learning email marketing

Email marketing means sending a message to people who handed you their address themselves, and its real value is that the list belongs to you and no algorithm can take it away. The hard part is not writing the email; it is getting the email into the inbox.

  • Lesson 6 of 8
  • Intermediate
  • Free, no signup

The email chain: from an address to a returning customer

Each link is written once and then goes out automatically.

  1. 1 The list

    a form on your own site, with a specific reason and a confirmation email

  2. 2 Welcome

    the thing you promised, plus two lines on what comes next

  3. 3 Two teaching emails

    useful to them even if they never buy

  4. 4 The offer

    one offer, one button, no fake countdown

  5. 5 Retention

    for someone who bought: a renewal reminder, or simply asking whether anything is wrong

This figure gives no interval between the links, because the right number depends on your work and any number here would be wrong for half the readers.

Last checked: Facts and tool names in this lesson are re-checked against their sources on this date.

How well does email work for an Iranian customer

Less well than the foreign books say. In the Iranian consumer market the slot those books give to email has been taken by Telegram and SMS: people here do not open an inbox to shop, they open it for work, for the bank and for invoices. If you run a clothing shop and expect a weekly newsletter to sell the way the American examples do, you will probably be disappointed.

But there is one thing Telegram and Instagram do not have, and it is what makes email worth the effort: the address is yours. A channel can be closed, a page can be limited, an algorithm can decide tomorrow to show your post to fewer people. The file of addresses sits on your own computer and nobody can take it from you. So look at email not as a sales channel but as insurance on your audience.

Three places where email genuinely works, and we say this from our own work rather than from market research: selling to businesses, where the person on the other side reads whatever you send in a work inbox; transactional messages such as invoices, verification codes and renewal reminders, which people are actually waiting for; and the returning customer, who has bought once and knows your name. For everything else a Telegram channel is often the better tool, and it is more honest to say so.

White glass envelopes passing three checkpoints lit red, green and blue with an almost empty spam bin at the end

Where the list comes from, and why a bought list wrecks it

From one place: a form on your own site with a specific reason to fill it in. "Subscribe to our newsletter" is not a reason. "The fourteen-point checklist to read before you order a website" is a reason, because the person knows what they get for their address. After they sign up, send a confirmation email and do not add the address to the main list until they have clicked its link; that makes the list smaller and better, and it keeps out typos and fake addresses.

Now why a bought list does not work, and the reason is technical rather than moral. Google requires anyone sending to Gmail to keep the spam rate reported in Postmaster Tools below 0.3%. That means out of every thousand emails you send, three people are allowed to press the "report spam" button. People who never gave you their address and do not know your name do exactly what anyone else would do: they report it. Those three fill up very fast.

And it is not the bought list that pays the fine, it is your domain. The bad history is recorded against the sending domain and IP, not against whoever sold you the list. After that your invoice and verification emails go to spam too, which means you have broken something that was never marketing in the first place. Ten thousand cheap addresses that work once, against a hundred addresses whose owners came on their own and can be written to for years: that is not a hard choice.

Why email lands in spam: SPF, DKIM and DMARC in plain words

Because the receiving server does not know you and has to decide from a handful of signals. Three of those signals you build yourself and publish in your domain's DNS.

SPF is a list of the servers allowed to send mail in your domain's name. The receiving server looks at which IP the message arrived from and whether that IP is on the list. DKIM is a cryptographic signature on the message itself: the public key is published in DNS and the receiver uses it to check that the text was not altered on the way. DMARC is a record telling the receiver what to do when those two do not line up with the sender address, and where to send the report. A DMARC policy can be to do nothing, to quarantine, or to reject.

Google has moved this out of the advice column and made it a requirement. Since 1 February 2024, everyone sending to Gmail accounts must set up SPF or DKIM for their sending domain, have valid forward and reverse DNS records (the PTR), use a TLS connection, and keep the spam rate below 0.3%. And anyone sending more than five thousand messages a day to Gmail must have both SPF and DKIM, must have DMARC as well (its policy may be set to none), and must offer one-click unsubscribe plus a clearly visible unsubscribe link inside marketing messages.

One point the guides rarely make: the PTR means that when someone looks up the sending server's IP in reverse, it must resolve to a name that resolves back to that same IP. That setting does not live in your domain's DNS; it lives in the panel of whoever gave you the server. If your mail keeps landing in spam for no visible reason and the three records above are correct, this is the place to look.

What an email passes through between your server and the inbox

Every checkpoint can send the message to spam, and all of them happen before anyone reads a word.

  1. 1

    Send

    the connection must use TLS

  2. 2

    Server identity

    valid forward and reverse DNS, the PTR

  3. 3

    SPF and DKIM

    is the IP allowed, is the signature intact

  4. 4

    DMARC alignment

    does the From domain line up with SPF or DKIM

  5. 5

    Sender history

    the reported spam rate must stay below 0.3%

The requirements on this path come from Google's own sender guidelines and speak only about Gmail. Other providers have their own, and we make no claim about them here.

The five emails almost every business needs

Not five emails a week. Five roles an email sequence plays, each written once and then sent automatically.

Welcome, right after the address is confirmed. Deliver the thing you promised, and say in two lines what you will send later and how often. This email has the best chance of being read, because the person pressed the button a minute ago. Two teaching emails a few days apart: something useful to the reader even if they never buy from you. This is where they decide whether to take you seriously.

The offer, after those two. Now that you have given something, you can say what you do and how to order. One offer, one button, no fake countdown. And retention, which is an email for someone who has already bought: news of a new capability, a renewal reminder, or simply a question asking whether anything is wrong. That last one gets the most replies and is the one almost nobody sends.

We do not put numbers on the intervals, because the right number depends on what you do and anything we write here would be wrong for half the readers. The rule we use ourselves is simpler: every email has to be worth the reader's time on its own, even if the next one never arrives. An email that fails that test does not get sent.

What to fix first, and what never to send at all

Before you send the first campaigndone once, felt forever

Do these before sending

  • set up SPF and DKIM for the sending domain
  • publish a DMARC record, even with a policy of none
  • check the server IP's PTR in the server panel
  • put a clearly visible unsubscribe link in the body
  • send a plain-text part alongside the HTML one
  • send to yourself first and look in the spam folder too

Never do these

  • buying a list of addresses or copying one from somewhere else
  • hiding the unsubscribe or making it hard
  • sending a whole campaign from the main domain on day one
  • a misleading subject line that does not match the body
  • making decisions on the basis of open rate

This list does not guarantee the inbox. Sender reputation is built over days and weeks, and no setting fixes it overnight.

What to measure, and why open rate is not a number you can lean on

Open rate is counted with a very small picture: the email body carries an invisible image, and when the mail app fetches it from the server, the server learns the email was opened. The trouble is that many mail apps fetch images ahead of time by themselves, even when the user has not opened anything.

Apple states this plainly: Mail Privacy Protection hides the user's IP address and "prevents senders from seeing if you have opened the email message they sent you". So every time your open rate moves up or down, it may have nothing to do with your subject line. How many of your readers use Apple Mail? We do not know, and we have seen no figure for the Iranian market; not knowing is itself the reason not to decide anything on this number.

Three things that can actually be leaned on: clicks on a link you tagged yourself and can follow in analytics; unsubscribes, the one honest piece of feedback you get for free; and whatever happened at the end of the chain, meaning the call, the order or the signup. If you only had room for one number, keep that last one.

The fast path, with AI

Writing five emails from scratch is exactly where the work usually stalls: one gets written and the rest never do. The fast move is not telling a model to write an email sequence, because then you get five unrelated pieces of ad copy. The fast move is to give it a short brief of your own offer first, then ask for a map of the chain with the role of each email, and only after approving that map, to take the texts one at a time. Three separate steps, not one large request.

  1. Write the brief in ten lines: what you sell, exactly who the customer is, what you gave in exchange for the address, and one thing your competitor does not have.
  2. Ask for the map of the chain: five rows, each row the role of one email and the single job that email has to do. Do not ask for copy yet, only the map.
  3. Correct the map yourself. This is where your judgement decides the result, and ten minutes here is cheaper than rewriting five drafts.
  4. Take the emails one at a time, and after each one say which sentence you did not like. A model works far better on the second pass than on the first.
  5. Fill in every number, price, date and name yourself. For mechanical work such as producing the plain-text versions, a cheap fast model of the Flash class is enough; for the brief and the map, use the stronger one. See the AI section of this site for today's pick.

Copy-ready recipe

Business: {what you sell, in one sentence}
Customer: {exactly who, not "everyone"}
What I gave in exchange for the email: {name the file, discount or guide}
One thing my competitor does not have: {something real and provable}
Tone I want: {for example plain, no slogans}

Do step one only, then stop:
Give me the map of a five-email chain as a table. Columns: number, the role of the email, the single job this email must do, one sentence that could be its subject line.

Rules:
- Write no number, percentage, price or date. Where a number belongs, write {number}.
- Make no claim about my product that is not in the brief above.
- No countdowns, no limited capacity, no manufactured urgency.
- Do not write the emails. The map only.

When I approve the map, write email one and stop.

Before you trust the output: Two things do not come out of this recipe and you have to put them in yourself. First, everything factual: prices, dates, specifications, names, and any sentence that will make a customer phone you if it is wrong. The blanks in the recipe are deliberate, so fill them in, and if the model wrote a number of its own, delete it. Second, none of these five emails reaches an inbox if SPF, DKIM and the PTR are not right; writing does not solve delivery, and the two jobs have nothing to do with each other.

AI in this kind of work

In email marketing, AI genuinely frees up time on two jobs and is dangerous on a third. It frees time on drafting the first version of a chain, and on producing several versions of a subject line so you can choose between them. It is dangerous the moment you ask it to talk about a product it has never seen, because that is where specifications and prices get invented, and email is the one channel that delivers such a sentence to the customer under your own domain name.

Tools that actually help

  • Claude When you tell it to write nothing outside the brief, it usually writes nothing and leaves the blank empty; for an email chain that matters more than fine prose. Iran is on neither of Anthropic's two supported-country lists, and we read that on Anthropic's own page.
  • Gemini Fast and cheap for producing ten versions of a subject line and then shortening them. Persian is on the supported-language list for the web app, but Iran is not on the supported-country list; both facts come from Google's own help pages.
  • NotebookLM If you already have a manual, a catalogue or a list of repeated customer questions, this tool answers only from those sources, which is what keeps a new claim from appearing. Google says it works in the same regions the Gemini app works in, and Iran is not on that list.

Where it backfires

The main risk here is not an SEO risk, it is a liability risk. A wrong sentence in an Instagram post can be edited; an email that has gone out does not come back, and if it carried a wrong price, specification or date, the customer keeps it as your promise. Models produce such sentences with complete confidence, and Anthropic keeps a page in its own documentation whose job is reducing exactly that, so the model maker does not deny the problem either.

The second risk is customer data. Your email list is other people's personal information, and pasting it into a chat box means sending it to a foreign company's server. Whether that text is used to train a model depends on the service, the plan and the settings, and each has its own page; Anthropic publishes one on this very question. The simple rule we use ourselves: no real name or address of any person goes into a prompt, and when the work needs an example, we invent one.

Sources: Anthropic docs: reduce hallucinations Anthropic: is my data used for model training Anthropic: supported countries and regions Google: where Gemini Apps are available Google: NotebookLM availability

Where this advice stops

Four things are not in this lesson. First, any number about results: not a usual open rate, not a click rate, not how many orders come out of a thousand emails. No such figure has been published for the Iranian market, and inventing one would be easy but not honest. Second, the requirements we quoted are Google's and specific to Gmail, and that page carries its own update date; Iranian and corporate providers have their own rules and we have not read them. Third, we make no claim about mail from foreign servers reaching Iranian email providers, because we have not measured it. And fourth, RGB does not sell an email marketing service; this lesson comes from running our own mail server, not from selling anyone a campaign.

From our own work

In July 2025 the signup verification codes for this site's own SEO tools sometimes arrived and sometimes did not. The plugin code was correct and wp_mail() was returning true, which proves nothing: that true only means the message reached Postfix. The cause was in /etc/postfix/main.cf: inet_protocols was set to all, so Postfix picked IPv4 or IPv6 at random for each delivery. This server's IPv6 address had no PTR record at all, and Gmail rejected every IPv6 attempt with 550-5.7.25 ... does not have a PTR record setup. Roughly half the mail to Gmail vanished, and on our side we saw no error.

The fix was two lines, inet_protocols = ipv4 and smtp_address_preference = ipv4, plus a full restart because that setting does not change on a reload. Both lines are still in place today. But the second half of the story is still open, and that is what makes it a lesson: the PTR for this server's IPv4 address is mail.rgbr.ir, with one extra letter, and that name does not resolve at all, while Postfix introduces itself as mail.rgb.ir. So the very requirement Google states, valid forward and reverse DNS records, is still not fully met on the server showing you this page. What we took from it is simple: a successful send report says nothing about delivery, and the only honest place to see the result is /var/log/mail.log.

Real follow-up questions

To start out, should I use an email service or send from my own server?

For a small list and transactional email, your own server with SPF and DKIM set up correctly is enough. For campaigns to a large list, a sending service makes life easier because it already carries the IP reputation, the unsubscribe handling and the bounce reports. Either way, the domain records remain your job.

How often should I send?

We do not know the number, and anyone who gives you one is not citing anything. The practical test is this: as long as every email is worth the reader's time on its own, the interval is right. The moment you start writing to fill a calendar, it has become too often.

My emails go to spam, where do I start?

Check the three domain records first, then the server IP's PTR, which lives in the server panel and not in the domain DNS. If all of them are right, the problem is sender reputation, and reputation repairs itself over some weeks by sending less, to addresses that actually want you. No setting fixes this overnight.

Keep learning