Aster and Rye sells loose leaf tea from a unit in Hastings to a list of 4,300 people. In June 2026 it sent one campaign, on a Tuesday morning, and it took £1,190. In the same month two automated sequences written once in 2024, which nobody touched at all, took £1,680 between them.
That comparison is the argument of this whole topic, and it is the one thing about email that almost nobody arranges their week around. The campaign is a morning's work every week and it stops the moment the morning is skipped. The flows are a fortnight of work, once, and they are still running.
The open rate is not a number any more
Opens have not been a real number since 2021, and the inflation is not evenly spread. Mail clients began fetching the images in a message before the person opened it, which is what an open is measured with, so an open now means that a machine somewhere decided to load a picture. The distortion is heaviest among whichever clients do the most pre-fetching, which means comparing the open rate of two segments compares their choice of mail app.
Clicks per delivered and revenue per delivered are what replaced the open rate, and both survive contact with reality because they need a person to do something. The June campaign went to 4,300 and was delivered to 4,180. It took 268 clicks and 41 orders. £1,190 from one campaign is £0.28 per delivered email, and that single figure is the one to carry forward, because it is comparable across every send the shop will ever make and the open rate is not comparable across two.
A flow is written once and sends itself; a campaign is a Tuesday
The welcome sequence at Aster and Rye is three messages, sent on days nought, three and nine after somebody subscribes. Around 310 people subscribe a month, the sequence turns 37 of them into first orders, and it made £1,070 in June without anybody logging in. The basket sequence is one message and it made £610. Neither has been edited since it was written.
Against those three lines sits £180 of refunds, which nobody files as an email number and which lands mostly on campaign orders, because a campaign sells to somebody who was not looking for anything. June kept £2,690, and £1,680 of it was earned by messages nobody in the building opened.
The arithmetic that follows is uncomfortable for anyone who spends a morning a week on the newsletter. A campaign earns once and its ceiling is the size of the list. A flow earns every month at the rate new people arrive, so its ceiling is the rate of subscription, and improving it improves every future month rather than one Tuesday. Which is why the emails that send themselves, forever is the cluster to read before the messages you send on purpose, even though every guide to this subject is arranged the other way round.
Whether it arrives is decided before anything is written
All of the above assumes the message reaches an inbox. That assumption became a great deal less safe in 2024. SPF, DKIM and DMARC have been required by the large mailbox providers since February 2024 for anyone sending in bulk, and a domain without all three is not filtered occasionally, it is refused predictably.
The second requirement from the same change is the one that gets missed: one-click unsubscribe is a header, not a link in the footer. A footer link has been in every message for twenty years and it does not satisfy the requirement, because the point of the header is that the mail client can offer the button itself without the reader ever opening the message. Aster and Rye discovered this in March 2025 when its delivery rate to one provider fell to almost nothing over four days, which is the whole subject of whether it arrives at all.
The one stage everything passes through
Every message written, through one gate, into everything it could do
Four things go in and five can come out, and all of it is decided by a stage that no amount of writing improves.
- The welcome sequenceThree messages, written once in 2024
- The basket sequenceOne message, and the shortest thing on the list
- The Tuesday campaignA morning a week, every week
- The receipts and dispatch notesThe most opened mail the shop sends, and the least written
- Clicks268 in June
- Orders41 from the campaign alone
- RepliesUncounted, and the only ones a person reads
- ForwardsInvisible, and the reason a list grows without advertising
- ComplaintsThe one that travels back up and narrows the waist
The waist is drawn as one gate and it is really three: the sending domain, the reputation of the address, and what this shop's own readers did last time.
Where this stalls: the list that is collected and never sent to
The stall in email is not writing and it is not tooling. It is that sending feels like an imposition, so the list gets built for two years and mailed four times, and every one of those sends is worse than the last for a mechanical reason. People do not remember subscribing eighteen months ago. Not remembering produces complaints rather than unsubscribes, complaints are the strongest negative signal a mailbox provider has, and the next send therefore reaches fewer inboxes than the one before it.
Sending rarely makes sending worse. That is the sentence that reverses most people's instinct, and it is why a monthly note that nobody agonises over outperforms a quarterly one that somebody did. The shop's four sends in 2024 each reached a smaller share of the list than the one before, and the list had grown the whole time.
What to open in 2026
The category to buy is an email service that will export the list as a file you can read, and the two obvious answers for a shop this size are Mailchimp and Klaviyo; both replaced sending to a hidden copy field from a normal mailbox, which stopped being viable the moment authentication became mandatory. Mailchimp has a free tier with a low contact ceiling. Klaviyo is priced on list size and is built around the flows rather than the campaigns, which is the right shape for a shop and the wrong shape for a newsletter.
The one non-negotiable is the export. A list you cannot download is a list you are renting, and the test takes two minutes: find the export button before you upload anything. Everything else about a provider can be changed later and that cannot.
Five clusters, and the order that is not the usual one
Permission comes first, because everything downstream is illegal or useless without it, and getting permission, legitimately is also where the complaint rate above is decided, two years before it shows up. Then flows, then deliverability, then campaigns. That order is deliberate and it is not the one most people follow.
Opens are broken, so what do you use closes it, and it exists as its own cluster because the answer is longer than one paragraph and every dashboard in this category still leads with the broken number.
Sideways, the list is fed by whatever the shop publishes, which is writing the things that get found and read, and the audience most likely to subscribe is the one already buying, which is selling online, and the details that decide whether it works. The rented alternative to all of it is building an audience on someone else's platform. It sits under only four of them are channels.
What Aster and Rye cannot answer is what the flows are actually worth, because the £1,680 counts every order that arrived after one of those messages, and some of those people had a kettle on and a favourite tea and were coming back regardless. Separating the two takes turning a flow off for a month, and nobody wants to be the person who turned off the thing that earns while sleeping.