Email marketing — where to start, and why open rates no longer tell you anything

A company weighing up its own mailing list is usually handed two things: a promise of the highest return in marketing, and a chart of open rates. The first cannot be checked, and the second measures something different today from what it measured five years ago — as both the maker of the email client and the firm publishing the metric will tell you.

This text is about what you need to know before the first send: which number stopped working, what conditions Gmail set, and how much a magnet for collecting addresses actually yields. The last part is our own measurement, because the only data we answer for is ours.

The metric that stopped working in 2021

Open tracking rests on one trick: a pixel-sized image inside the message, fetched from the sender's server. Fetching the image means "somebody opened it". An entire industry built its headline metric on that.

In September 2021 Apple introduced Mail Privacy Protection. Apple describes it in its own words:

> Mail Privacy Protection hides your IP address, so senders can't use it to determine your location and create a profile of your mail activity.

and, directly about our metric:

> Mail Privacy Protection also prevents senders from knowing whether you've opened their email.

The mechanism is that Apple's mail fetches images in advance, whether or not anyone opened the message. To a sending platform that looks like an open. Every such message lands in the statistics as read, though it may never have been displayed.

What makes this worth stating is that the other side confirms it. GetResponse, publishing its benchmark, warns against its own number — it writes that the rise in engagement "can partially be attributed to the growing popularity of Apple's devices, which distort the data with its auto-opens". The publisher of the metric says the metric is inflated.

The practical conclusion is blunt: the open rate is fit today for comparing messages with each other in the same period, and for nothing else. It is not fit as a target, not fit for year-on-year comparison, and not fit as evidence of effectiveness — because part of the rise comes from devices rather than people.

What the numbers actually are

With open rates set aside, it is worth knowing what remains. The most useful public reference point comes from the GetResponse report, because it is one of the few that states its sample and breaks the data out by country.

The sample: 4.4 billion messages sent by GetResponse customers in 2023, counted for senders with at least 500 contacts, broken out across 24 countries and 18 industries. That is a level of detail most circulating "industry averages" do not have.

Global averages: open rate 39.64%, click-through 3.25%, unsubscribes 0.15%. The report gives a row for each of those 24 countries, and yours will be among them — worth looking up rather than taking the global figure as local.

Two caveats without which these values mean less than they appear to. They are the data of one platform's customers, not of the whole market. And they describe senders who already have a list and send to it — which means they say nothing about how hard that list is to build. That is a separate question and we return to it below.

What Gmail has required since February 2024

This is the part nobody mentions to a company before it starts sending, and it decides whether the messages arrive at all.

Since 1 February 2024 Google requires four things of senders of 5,000 messages a day or more:

• SPF and DKIM — authentication of the domain the mail leaves from. Without it, a message looks to a filter like impersonation.
• DMARC — and, notably, the policy may be set to `none`. What is required is the record, not strict enforcement. That is a far milder condition than it is usually presented to be.
• One-click unsubscribe, plus a clearly visible unsubscribe link in the message body. Not a form, not a login, not "reply to this message".
• A spam complaint rate below 0.30%, measured in Postmaster Tools.

The 5,000-a-day threshold sounds high and most small companies will never touch it — but three of those four things are done once and hold forever, and the fourth is simply good practice. Treating them as obligatory from the first message saves coming back to the subject when the list grows.

The last point has a further practical meaning: 0.30% is very little. On a list of two thousand, six spam complaints breach it. That is a real reason not to buy address lists and not to add people because they once sent an enquiry.

Where addresses come from — our own magnet, measured

Here we reach the question benchmarks do not cover: how much has to go in before somebody leaves an address at all.

We do not run a newsletter, but we do run a magnet, and we have measured it. Our cost calculators are a lead magnet in the best possible form — a tool that genuinely calculates something and whose result has standalone value even to somebody who never orders from us. At the end there is an offer: we will email you this result.

Three figures from half a year, given as shares:

• 84% of those who started calculating finished. The tool itself works — people do not drop out midway.
• 9% of those who finished opened the form offering to send the result by email.
• About 2.4% of those who finished actually left an address — roughly one person in forty.

The caveat belongs beside these numbers, not beneath them. The last step rests on a handful of events. At that size it is an order of magnitude, not a measurement — the true value could be twice as high or twice as low and we have no way to settle it. We publish it anyway, because an order of magnitude is still information and this industry rarely offers even that. A second caveat: analytics sees only people who accepted cookies, so each of these shares is a floor.

The conclusion we draw, worth more than the number itself: building a list is slow even with a good magnet. A tool that genuinely calculates money yields a few per cent of addresses from people already engaged. Any plan assuming "we will collect a thousand addresses this quarter" should be worked backwards through that fraction before anyone approves it.

What it costs once the list grows

A mailing platform's price list is usually read once, at sign-up, and what sticks is the lowest figure in the table. That is the moment to look one column further, because the price of these tools does not rise in proportion to the list.

We collected the price lists of six suppliers on 25 and 26 August 2026 — ActiveCampaign, Brevo, FreshMail, GetResponse, Mailchimp and MailerLite — as part of our own research into tool costs. Thirty-five observations, each with the plan, currency, billing period and date of collection.

The sharpest case in the set: the Starter plan at GetResponse costs 33.7 times more at ten thousand contacts than at one thousand — the same plan, a tenfold larger list, a bill more than thirty times higher. For comparison, ActiveCampaign over the same step rises 9.9 times, almost exactly in proportion to the list.

The amounts themselves were listed in Polish zloty and US dollars on the Polish market, so we give the multiplier rather than the price: the multiplier is the finding and it does not depend on currency. Check your own market's list prices — but check them at the second threshold, not the first.

This is not an accusation against any supplier; billing models differ and each is entitled to price as it wishes. It is, however, an argument for reading the second tier of the price list before you start collecting addresses. A growing list is a success; a bill growing three times faster than the list can cancel that success in a quarter.

Practically: work out the cost not for the list you have but for the list you plan to have in a year — and do it with two suppliers, not one. The difference between them is often larger than the difference between plans at the same one.

What works as a magnet and what does not

The rule follows directly from the numbers above: what you give in exchange for an address must have value in itself — including to somebody who will never buy from you.

What works: a calculation for a specific situation (cost, timeline, risk), a comparison of prices or data that exists nowhere else, a document template ready to fill in, and a checklist to tick off before a decision. The common factor: the person receives something ready to use, not an announcement.

What does not work: "subscribe to our newsletter" without saying what will be in it and how often. A discount for somebody who does not yet know whether they want to buy — it attracts discount hunters who then raise your unsubscribe rate. And an ebook about everything, whose value is not obvious to the reader until they have read it.

There is one more condition, easy to miss: the magnet has to stand where the person is already engaged. Ours works because the offer appears after the calculation, not before it. The same offer in a pop-up on arrival reaches somebody who does not yet know whether they want to stay.

Consent, unsubscribing and data

Without alarmism and without pretending to be legal advice, because it is not. Three things worth having in order from the first send:

Consent must be freely given and separate. An address left in order to receive a calculation is consent to receive that calculation. It is not automatically consent to a weekly bulletin — that is a second matter and a second decision by the person. Mixing them in one checkbox is a common mistake and the one that most quickly ends in a spam complaint.

Unsubscribing must be immediate and take one click. That is now not only good practice but a Gmail requirement, as above. Hiding the link, requiring a login or asking somebody to "confirm once more" turns a person who merely wanted to leave into a person who clicks "this is spam" — and that is the same 0.30% rate.

You have to know where every address came from. Not for the sake of a regulator, but because a list without that knowledge is useless the first time deliverability goes wrong. Record the date, the source and precisely what the person agreed to.

When email is the wrong channel

A text about email marketing that never says who should not do it is an advertisement, not a guide. Three situations where a mailing list is the wrong answer:

When you have a few dozen customers and know them by name. At that scale a bulk sending tool solves no problem an ordinary inbox does not. The cost of a platform, the time to configure authentication and the upkeep of a list make sense only once writing to each person individually is no longer possible — and not before.

When the sale is one-off and rare. A building firm fitting a bathroom for somebody once a decade has nothing to communicate cyclically. The list will age faster than it produces a single order, and the unsubscribe rate will tell you so after three sends.

When you have nothing to give. This is the most common case and the hardest to admit. If the only thing that comes to mind as the content of a regular message is "company news", the problem is not the channel but the absence of material — and email delivers material, it does not create it. Better to start with one thing genuinely worth sending and build a reason to collect addresses out of that.

The honest order is therefore the reverse of the way it is usually sold: first something of value, then a magnet, then a list, and a tool last. Starting with the tool is the most expensive way to discover you have nothing to send.

What to measure instead of opens

Since the first section takes away your headline metric, you are owed a replacement.

Clicks. They require a deliberate human action and no privacy mechanism generates them. This is the closest thing today to what open rates were meant to be.

Replies. An underrated metric, particularly in B2B. A message somebody answers is a message that worked, and replies improve your sender reputation with the mail provider as a side effect.

Conversions on your site, from tagged links. Without parameters in the addresses, traffic from mail falls into a "direct" bucket and a month later you cannot tell the send that worked from the one that did not. How to set that up is covered in the article on analytics tools, and what to do with that traffic afterwards in the one on conversion rate.

Unsubscribes and spam complaints. Not as a source of shame but as an early warning system. A jump in unsubscribes after one send says more about that send than any open rate.

What is not worth doing: comparing your own open rates year on year, taking industry averages as a target, and drawing conclusions from a single send.

What we do not have, and do not pretend to

We end as we began — with what can be checked.

We do not run a newsletter. We collect addresses through calculator results, we send transactional messages, and we have measured our own magnet — and that is the only thing we can speak of as experience. We have no data of our own on open rates, on the click-through of a message series, or on how a list behaves after a year of sending. Had we quoted such figures, they would have been borrowed.

We say this plainly, because an article about a metric the whole industry quotes without qualification cannot itself stand on borrowed authority. Everything above that is ours is labelled as ours and given in shares; everything that is somebody else's carries a publisher, a sample and a date.

If you take one thing from this: before you build a plan on open rates, check whether they measure people at all. Since 2021 they largely do not, and both the maker of the mail client and the firm publishing the metric say so.

Sources, all checked directly with their publishers in September 2026:

• Apple, "Use Mail Privacy Protection on iPhone" — the mechanism that hides the IP address and the fact of opening.
• GetResponse, "Email Marketing Benchmarks" — 4.4 billion messages in 2023, senders with 500 contacts or more, 24 countries, quoted together with the publisher's caveat.
• Google, "Email sender guidelines" — requirements for bulk senders in force since 1 February 2024.

Figures about our own magnet come from GA4 over six months, cover only users who accepted cookies, and are given exclusively in shares. The price observations were collected on the Polish market on 25–26 August 2026 and are given as multipliers rather than amounts.