Sales Email Subject Lines: The Metric That Stopped Working

A sales email subject line has one job, and it is smaller than the advice around it suggests. It has to get the first line of the body read. Not opened, not admired, not remembered. Read the next sentence, which is where the actual decision happens.

Almost every rule people argue about follows from that, and the arguing has outlasted the measurement that used to settle it. The subject line is the most-optimised element in outbound and the one whose success metric quietly stopped working.

What the reader is actually looking at

Nobody sees a subject line on its own. They see an inbox row, and the row has three parts competing for the same half-second: the sender name, the subject, and the preview text lifted from the top of the body.

The sender name is read first and it is doing more work than the subject. An unfamiliar name at an unfamiliar company sets the frame before a single word of the subject registers, which is why a clever subject on an unrecognised sender reads as a trick rather than as intrigue.

The preview text is the part most people forget they are writing. Mail clients pull it straight from the opening of the message, so a first line that begins with a greeting spends the preview on the word "Hi". The subject and the preview are read as one phrase, and they can be written as one phrase.

Sender nameRead first

• Sets recognised or not recognised before the subject registers
• Cannot be fixed by the subject line
• A mismatch between name and subject reads as a trick

SubjectRead second

• Truncates hard on mobile
• Has to survive being read as a fragment
• Its only job is to earn the first line

Preview textRead as part of the subject

• Pulled from the opening words of the body
• Wasted on a greeting by default
• Written deliberately, it doubles the subject's length

The measurement problem underneath the whole debate

Subject lines have always been optimised on open rate, and open rate is measured by a tracking pixel: an invisible image the sender embeds, which records a load when the reader's client fetches it.

Apple's own support documentation describes the mechanism from the other side. Its page on reducing junk mail tells readers that spammers use the email image-loading feature to determine whether an account is active, and instructs them to turn on Protect Mail Activity so images stop loading automatically. Apple is describing, accurately, the exact instrument that produces an open rate. When a mail provider preloads or blocks those images on the reader's behalf, the sender's counter fires without a human involved, or does not fire when one was.

The consequence is not that open rate became slightly noisier. It became a mixture of two different populations, human readers and machine prefetches, in a ratio the sender cannot see and which varies with the mail clients on the list. Comparing this month's open rate to last year's compares two different instruments.

Which leaves a straightforward question about what to optimise instead, and only one answer that survives it: the reply. A reply is a human decision, it is what the campaign exists to produce, and no privacy feature generates one by accident. The published open-rate literature is still worth reading for its measurement caveats rather than its targets, and the cold email open rate benchmarks page collects both.

Two practical consequences follow. Any subject-line test decided on opens alone is decided on a corrupted instrument. And a subject line that lifts opens while depressing replies has done damage, which is the exact failure mode of curiosity-gap and fake-familiarity lines.

The rules that survive contact with a reply rate

The subject must be honest about what the body contains. A line that promises something the message does not deliver buys an open and spends the reply. This is the rule that curiosity-gap advice violates by design, and the reason it survives is that it looks like it works when you are grading on opens.

Length matters because of truncation, not because short is virtuous. A subject is read as whatever survives the cut on a phone. Write the line so that its first few words carry the meaning, then check what remains when it is truncated.

Specificity beats cleverness, and the specific thing must be true. A subject naming something real about the recipient's company works because it could not have been sent to anyone else. That property is the whole of it, and it disappears the moment the specific detail is a merge field with a generic sentence around it.

Fake familiarity is the most expensive shortcut in the category. "Re:" on a first message, "following up" when nothing preceded, "as discussed" when nothing was. These lift the open, produce a distinct kind of irritated non-reply, and cost the sender the one thing outbound cannot buy back.

The subject is written last. You cannot title a message before you know its premise, and the ordering matters more than any individual rule: our own process piece, how to write a cold email, puts the subject at the end for exactly this reason.

Lower case, no punctuation and other house superstitions are worth testing and not worth believing. They are cheap to try, they occasionally matter, and none of them will rescue a message whose premise is weak. Treat them as the last percent rather than the first.

Before the campaign goes out

• ✓Yes: The subject is true about what the body actually says
• ✓Yes: The meaning survives truncation on a phone
• ✓Yes: The preview text was written deliberately rather than inherited from a greeting
• ✓Yes: Any specific detail in the line is verified rather than generated
• ✕No: Fake thread markers such as Re: or following up on a first message
• ✕No: A curiosity gap the body does not close
• ✕No: Deciding the test on opens alone

Why the swipe files do not transfer

Every list of the best sales email subject lines is a list of outputs. Each line worked once, for one offer, sent to one audience, by one sender, at one moment. What made it work was the fit between those things, and the fit is the part the list cannot carry.

A subject line is also the shortest element in the message and the one most exposed to repetition. A formula that circulates widely enough stops working through familiarity alone, and the most-copied lines are by definition the most familiar to the people receiving them. The half-life of a good template is set by how many people copy it.

The transferable part of a swipe file is the structural observation: which lines are statements, which are questions, which name a specific thing, which reference a time. Reading a list for its structure and writing your own line inside the structure that fits your premise is a different exercise from pasting the line.

Where our position changes the advice

We run one message per campaign, with no bumps and no thread replies, and re-entry is a new campaign built on a new premise rather than a reminder of the old one.

That has a direct effect on the subject line, and it is the reason a lot of standard advice does not apply here. Where a sequence exists, the first subject line can afford to be a probe, because a later message can recover from it. With one message, the subject carries the entire campaign. There is no second line to correct a misfire and no thread for a follow-up to sit under, so the subject and the first sentence have to do all of the work, once.

The second consequence is about what a subject line can honestly say. One message per campaign means the thread-reply device and the follow-up device are not available to us at all, because we do not send those messages. That removes an entire category of high-performing-on-opens subject lines from the toolkit, and it is a cost of the position rather than a free win.

The third is that testing gets simpler and slower. One message per campaign means a subject-line comparison is a comparison between campaigns rather than between steps, and the honest read needs volume before it means anything.

Here is illustrative arithmetic showing why, invented for the purpose and not a measurement of any campaign. The figures below are invented illustrations. Suppose two subject lines are tested across 400 recipients each, and one produces 8 replies while the other produces 12. That looks like a 50% improvement. At those counts, a difference of four replies is well inside the range that ordinary variation produces, and running the same test again could reverse it. Subject-line differences are usually small, and small differences need larger populations than most campaigns have.

Which is the honest summary of the whole topic. The subject line is worth getting right and it is rarely the reason a campaign works or fails. The list and the premise decide that, and the line on top can only fail to get them read. What a usable target definition contains is in ideal customer profile.

The short version

The subject line's only job is to get the first line read, and it is read alongside the sender name and the preview text rather than on its own. Write the preview deliberately and you have doubled the space.

Optimise on replies rather than opens. Open rate is produced by a tracking pixel, and mail providers now load or block those images on the reader's behalf, so the metric mixes human readers with machine fetches in a proportion nobody can see.

Keep the line honest, specific and short enough to survive truncation, and refuse the fake-familiarity devices that trade a reply for an open.

Read swipe files for structure rather than for lines, since the fit between offer, audience and moment is what made any given line work and is the part a list cannot transfer. If you would rather see the whole motion built against your own market than assemble it line by line, we will build a campaign and show the reasoning.

Platform behaviour above was verified against Apple's own support documentation as fetched on 16 August 2026. Verify current terms with the vendor before relying on them.

Sources: Apple Support, Report and reduce spam in iCloud Mail