Subject lines that get opened without sounding like spam

The subject line is not read in isolation and it is not read carefully. It is scanned in a list, alongside a sender name, a timestamp and a fragment of preview text, in a few hundred milliseconds during which the reader is deciding whether this row is worth a tap.

Which means the sender name is doing more work than the subject line. A recognised sender gets read with the benefit of the doubt; an unrecognised one has to earn attention from the subject alone. If your from-name is inconsistent between sends, or it is a person nobody has heard of rather than the business the subscriber signed up to, you have made the subject line's job considerably harder before writing a word of it.

The third piece of copy is the preheader — the preview text most clients show after the subject. It is frequently left as whatever the template happens to start with, which is often 'View this email in your browser'. Treating it as a second line of the subject, extending rather than repeating it, is close to free.

Where a subject gets cut depends on the client, the device, the orientation and whether the reader has increased their system font size. There is no character count that is universally safe, and any specific number presented as a rule is a simplification of a pixel-width problem.

The useful version of the advice is directional: assume the narrowest case, because most mail is read on a phone. Roughly the first 40 to 50 characters are what you can rely on being visible in a portrait phone inbox, and everything after that is a bonus you should not depend on. Front-load accordingly — the specific noun, the date, the order number, the thing that makes this message different from the last one.

The common failure is a subject that spends its visible portion on setup. 'We are excited to announce that our new...' has used the entire safe budget before reaching any information. Cutting the first three words is usually the whole edit, and the fastest way to see the problem is to send yourself the draft and look at it on a real phone.

The folklore says certain words trigger filters: free, urgent, guaranteed, act now. As a description of how modern filtering works, this is largely obsolete. Providers weigh sender reputation, authentication, complaint rates and engagement history far more heavily than the presence of particular vocabulary, and a trusted sender can use the word 'free' in a subject without consequence.

But the advice survives for a different and better reason. All-capital letters, rows of exclamation marks and manufactured urgency do not primarily upset the filter — they upset the reader. A reader who feels shouted at is measurably more likely to hit the spam button, and the complaint rate is a signal the filter genuinely does weigh. The tactic fails, just not by the route it is usually blamed on.

Getting the causation right matters, because it changes what you optimise. Running a subject line through a 'spam word checker' addresses a mechanism that barely exists. Not sounding like a stranger shouting addresses the one that does.

The reader is not evaluating your writing. They are answering one question — is this relevant to me right now — and a subject line that is specific answers it faster than one that is clever.

Compare a subject that names the thing to one that hints at it. 'Your order shipped, arriving Thursday' tells the reader exactly what has changed. 'Big news from us' asks them to open in order to find out whether they care, which is a request rather than an offer. The curiosity gap is not useless, but it spends trust: it works the first time, and by the third time the reader has learned that opening your mail does not reliably reward them.

Specificity also protects you from the worst outcome, which is not a low open rate but an open followed by disappointment. A vague subject that oversells produces a tap and then a reader who feels misled — and that reader is the one who unsubscribes or complains. A subject that describes the contents plainly attracts fewer, better opens, which is the trade you want.

Inserting a first name into a subject line is not personalisation in any sense the reader experiences. It is a merge field, it is recognisably a merge field, and by now most readers have seen enough of it to read it as a marketing convention rather than as familiarity.

The version that works is relevance, not naming: the class someone attended, the product they bought, the city they booked in, the appointment they have on Friday. That is personalisation the reader could not mistake for automation, because it depends on knowing something specific about them.

Merge fields also fail loudly. An empty field produces 'Hi ,' and a badly imported one produces 'Hi FIRSTNAME' or somebody's full email address, and both are visible to the entire segment at once. If you use tokens, set a fallback for every one of them and check the fallback renders correctly. The failure mode is not a slightly worse subject line; it is a subject line that announces the mail was assembled carelessly.

The standard method — split the list, compare open rates, keep the winner — has been undermined by the thing it measures. Because Apple Mail's Privacy Protection preloads images before a recipient acts, reported opens include events no human caused, and the pollution is not evenly distributed between variants in a way you can correct for. A subject line A/B test scored on opens is now measuring a quantity that is partly independent of the subject line.

So score on what the reader did afterwards: clicks, replies, bookings, purchases. These are further down the funnel and therefore noisier, which brings the second discipline — volume. A difference of a few percent on a few hundred recipients is indistinguishable from chance, and treating it as a finding is how teams accumulate confident rules that are not true.

Test one variable at a time, and hold results loosely. A subject that wins on one send to one segment is evidence about that send, not a principle. Rules worth keeping are the ones that survive repetition.