- Judge subject lines by reply rate, because open rates now include automatic pixel loads that no human triggered.
- Write subjects as short, specific noun phrases that look like a note from a colleague, and drop the fake Re: and Fwd: prefixes.
- Read the subject, the preview and the first line together as one three-step filter, and fix whichever step is weakest.
For years, cold email subject lines were the most tested element in outreach, and for years that testing was measured by opens. Then Apple began pre-loading the tracking pixels in Mail on behalf of its users, and Gmail started routing images through its own proxy. Now a large share of “opens” are the mail client fetching a pixel, not a person reading anything. An open-rate test on two subject lines is partly a test of how often a proxy ran, and that tells you very little about which subject a human preferred.
That’s the first thing to change. A cold subject line has one job that is measurable: getting a specific person to read the first line. The best proxy for that job is a reply, and the reply is also the outcome you actually want. This article covers why opens stopped being reliable, the subject patterns that earn replies in practice, the ones that backfire, and how to test subject lines when your dashboard is lying to you.
Why open rates stopped measuring subject lines
The mechanism matters, because it explains why an open-rate win can be fake. When Apple introduced Mail Privacy Protection, it began fetching remote content, including the invisible image that tracking systems use to detect an open, for every message delivered to an Apple Mail account. That fetch happens whether or not the recipient ever looks at the email. Gmail’s image proxy does something similar for images it serves. The result is that opens are inflated, sometimes dramatically, and they’re inflated unevenly across audiences that use different clients.
The consequence for subject-line testing is severe. If Variant A goes to a list heavy with Apple Mail users and Variant B goes to a list heavy with Outlook users, the open-rate gap may be entirely a client gap. Even a fair split can produce a misleading winner, because the proxy fetch is not randomly distributed across people who would have opened the email on their own.
Replies don’t have this problem. A reply requires a human to write something. It’s noisier than it sounds, because some replies are “not interested” and some are out-of-office autoresponders. But it’s far more honest than an open, and most of the noise can be filtered with a few rules.
| Metric | What it measures now | Reliability for subject lines |
|---|---|---|
| Open rate | Pixel loads, including proxies, plus some human opens | Low, and uneven across clients |
| Click rate | Clicks on tracked links, from people who chose to click | Moderate, but only if the email has a link |
| Positive reply rate | People who wrote something that signals interest | High, but needs enough volume |
| Total reply rate | Any human reply, including negatives | Good for direction, count the negatives separately |
The practical rule is simple. Use reply rate as the score, use positive reply rate as the tiebreaker, and treat open rate as a rough hint about deliverability rather than about the subject.
The job of a cold subject line
A subject line is the first of three filters. The subject decides whether the message gets opened at all. The preview text decides whether it reads like a person. The first line decides whether the reader has a reason to answer. Each filter operates on a few seconds of attention, and each one can only pass a reader forward. None of them can rescue a weak step before it.
That changes what cold email subject lines should do. It doesn’t need to be clever, and it doesn’t need to sell. It needs to look like it came from someone who knows the recipient, about something the recipient already deals with. A subject that sounds like an internal note passes the first filter. A subject that sounds like a newsletter fails it, however good the copy.
The standard I use is this: would a person who knows the recipient plausibly send this subject to them? If the answer is yes, the subject is doing its job. If the answer is that only a marketing team would write it, the subject is working against you.
Cold email subject lines that earn replies
The patterns below are the ones I see working most reliably. Each one resembles a message between two people rather than an announcement from a company.
| Pattern | Example | Why it works | Where it fails |
|---|---|---|---|
| Specific noun phrase | “routing for new reps” | Reads like a topic a colleague would raise | Too vague to mean anything to the recipient |
| Their system or process | “your month-end close” | Signals relevance without a pitch | Wrong if you guessed their process |
| Lowercase question | “still doing reorders by hand?” | Easy to answer, sounds conversational | Yes-or-no questions with no stake attached |
| Company-plus-signal | “acme’s two new SDR roles” | Shows you looked, and it’s checkable | Stale signals read as a canned line |
Most strong subjects share three traits. They’re short, usually under five words. They name something specific to the recipient’s world rather than to yours. And they avoid capitalization and punctuation that signals a campaign. Lowercase isn’t magic, but it reads as a note typed quickly, which is the register you’re aiming for.
Notice what’s missing. There are no product names in the subject, no percentages, no “introducing” or “exclusive,” and no emoji. A product name in a subject tells the reader you’re selling before they’ve decided whether they care. A number is a claim they’ll want to check. Those choices are defensible in marketing email, where the reader has opted in. In cold email, they read as the opposite of a conversation.
Templates and subjects should match. If your first line is a question about a hiring signal, the subject should reference the same signal or the same problem. A subject that points to one topic and an opener that opens on another is the most common mismatch we see, and it is one of the fastest ways to lose a reply. We covered the structure of the opening line in our piece on cold email templates that actually get replies.
Patterns that backfire
Some subject habits look effective and hurt you. The first is the fake reply or forward. Writing “Re:” or “Fwd:” in front of a subject on a first email makes the message look like a continuation of a thread the recipient has already seen. It’s a bait tactic, and it works in the sense that it gets opened, then it fails in the sense that the recipient realizes there was no earlier thread. It also crosses a line. US law under CAN-SPAM prohibits subject lines that mislead recipients about the content of a message, and a fake thread is a clear example. My position is firm: don’t do it.
The second is the headline. “Introducing the leading AI platform for outbound teams” reads like a press release, because it is one. Headline subjects describe the sender’s pride rather than the recipient’s problem.
The third is the curiosity gap with no payoff. “Did you see this?” or “quick thought” gets a click or an open, and then the first line has to deliver something worth the wait. If it doesn’t, the reader feels tricked, and a tricked reader doesn’t reply. Curiosity can work, but only when the first line pays it off immediately.
The fourth is the name drop without a relationship. Putting a famous client’s name in the subject when you have no connection to the recipient is a credibility problem the recipient will notice in seconds. If you have a real reference, use it in the body, where it can be framed honestly.
The fifth is the forced emoji or special character. These are often read by filters and by people as signs of mass mail, and they rarely add anything that a plain subject wouldn’t do better.
No fake threads: a "Re:" or "Fwd:" on a first cold email misleads the recipient about the message's history. It can cost you the reply and the sender's credibility, and it is the kind of misleading subject that CAN-SPAM was written to prohibit.
Testing subject lines when opens lie
If you can’t trust opens, the testing method for cold email subject lines has to change. The core principle from testing templates still applies: change one variable, keep the rest fixed, and wait for enough volume. The difference is the metric. Score each variant by reply rate, and count negative replies separately.
Volume is the constraint. If your baseline reply rate is around 2 percent and you want to detect a move to 3 percent, you need roughly 3,800 sends per variant. Most subject-line tests are run with a few hundred sends and declared winners on a handful of replies. Those results are close to random. If you can’t reach that volume, test direction instead. Run the two most different subjects you can think of, see which one draws more positive replies over a few weeks, and use that to guide the next round rather than as a verdict.
Read the replies. This is the most underused method in cold email. When a subject works, replies often quote it or respond to it directly, with lines like “Yes, we’re still handling reorders by hand.” When a subject fails, replies tend to ask what you mean, or they don’t exist. A sample of twenty replies read by a person will teach you more about your subject lines than a week of dashboard numbers.
A useful trick is to keep a holdout segment that receives your current best subject with no changes. If the holdout’s reply rate drifts down over time, the problem may be the list, the deliverability, or the season rather than the subject. Without a holdout, you’ll blame the subject for problems it didn’t cause.
A worked example: three subjects, one email
Here is one email with three subject candidates. The body is a single question about a manual process at a company that has recently grown its finance team. The first line names the signal and asks whether the close is still manual.
The first candidate is “quick question about your close”. It’s clean and plausible, and it passes the test of sounding like a colleague. It’s also generic, which means the recipient has to read the first line to learn why you’re writing. Its weakness is that it could be sent to any finance team.
The second candidate is “your month-end close, still manual?” It names the specific process and ends with the question the body asks. It’s a little more committal, and it gives the recipient the topic and the ask in one glance. This is the one I’d test first against the generic version, because the gap between them measures how much specificity helps.
The third candidate is “Acme’s close process: a 5-minute fix?” It’s the one I’d avoid. The number is a promise the body has to keep, the company name in the subject feels like a template token, and “fix” is a pitch word. It may get more opens, and some of those opens will read the first line and feel misled. The reply rate is the test that will tell you that.
Read the subject and first line together: Read them aloud as one message. If they sound like two different people wrote them, the subject is promising something the body won't deliver.
Wrap-up
Subject lines are a three-step filter, and their job is narrow: get a specific person to read the first line. Judge them by reply rate, because open rates now carry pixel noise you can’t remove. Write short, specific subjects that look like a colleague’s note, make sure they match your opening line, and drop the tricks, especially the fake thread. Test with enough volume to mean something, and read the replies yourself, since they’ll show you which words the recipients recognized.
If you’re running this from your browser, OutboundLeadPro’s AI sequences with A/Z testing let you compare variants and read reply rates per variant in one place, so the subject-line decision is made on the metric that counts. You can see how the pieces fit on the features page. For the wider sequence, our guide to cold email sequences that get replies covers the follow-up structure that comes after the first open.
Keep reading
Cold Email A/B Testing: Why Most Tests Prove Nothing
Cold email A/B testing is often too small to mean anything. Here is the sample-size math at real reply rates, and how to read the results.
Daniel Khoury
September 19, 20263 min readHow to avoid spam folder placement: 12 causes and fixes
Want to avoid spam folder placement? These 12 causes, from authentication to list quality to complaints, each come with a concrete fix you can check.
Daniel Khoury
September 24, 20263 min readHow to build a lead list that stays clean after the first send
Build a lead list around a real ICP, with stage gates that reject bad rows before they cause bounces and complaints. Includes a worked example.
Daniel Khoury
October 7, 20263 min readSPF, DKIM and DMARC explained for cold email
The three email authentication records decide whether you reach the inbox — and most cold outreach fails because one is missing or wrong. Here's each one in plain English.
The OutboundLeadPro Team
October 11, 20262 min read