Most subject-line advice is measured on open rates, and in 2026 open rates mostly measure machines. This piece is about the other number, replies. It's deliberately modest: what's worth doing, what to avoid, and how to tell whether a change worked.
A note on numbers. An earlier version of this post quoted reply-rate lifts from "Outsolvi's send-base data". We don't publish statistics about our customers' email, and we couldn't stand behind those figures, so they're gone.
Why opens are the wrong scoreboard
Apple Mail accounted for 62 percent of the opens Litmus measured in July 2026,[1] and Apple Mail Privacy Protection fetches the tracking image when an email arrives, whether or not anyone reads it. Security scanners at large companies open links and images to check them. At large corporate domains, Omnivery found 80 to 95 percent of clicks came from bots.[2]
So a subject line that "wins" on opens may simply have reached more Apple Mail users. A reply needs a person. Judge subject lines by replies, and if you track opens, look at the ones graded as a person rather than the raw count.
What to write
Something specific to the recipient's work. "Pricing for your Q3 rollout" says more than "Pricing question". "Your team's Salesforce migration" says more than "Quick question about your stack". Specific subjects read like one person writing to another, because they are.
The important words first. Phones cut long subject lines off, so put the part that matters at the start. Four to seven words is a reasonable rule of thumb, not a law.
Sentence case. "Quick thought on Acme's expansion" reads like a colleague. "Quick Thought On Acme's Expansion" reads like a newsletter.
What's actually inside. The subject is a promise about the email. If the body doesn't keep it, the reader notices, and the reply doesn't come.
Personalization that shows you did the work. A first name in the subject is easy to generate and easy to ignore. A reference to the recipient's team, product or a recent change is harder to fake, which is why it works.
Follow-ups that say what's new. "Following up" tells the reader nothing. "The case study you asked about" or "Pricing for three seats" gives them a reason to open this one.
What to avoid
False urgency. "Last chance" and "Urgent" on a first cold email are rarely true, and readers learn to discount them.
Fake Re: and Fwd: prefixes. They make a cold email look like part of a conversation that never happened. Even when that wins an open, the reader feels tricked the moment they see the body. In the US, the FTC's guidance on the CAN-SPAM Act says a subject line must accurately reflect the content of the message.[3]
All caps and piles of punctuation. They read as marketing at best and spam at worst.
Emoji, usually. In a sales email they tend to read as marketing. If you want to try one, test it.
Examples for common situations
These aren't templates to paste. They show the pattern: say what the email is about, in words that fit this one reader.
- A first email to a named account. "Idea for Acme's onboarding emails" or "Question about your Q4 hiring plan". Both name something the reader owns, and neither pretends to know them.
- After a call. "Notes from Tuesday, plus pricing". The reader knows what's inside, and why it matters to them, before they open it.
- Sending something they asked for. "The security questionnaire you asked for". It answers the question for them, and it's easy to find again later.
- A warm introduction. "Priya suggested we talk". It only works when it's true, and when it is, it's the strongest line on this list.
- A deal that went quiet. "Still on for the pilot in March?" A specific question is easier to answer than "Checking in", and a no is still useful to you.
- The last email in a sequence. "Should I close this out?" It gives the reader an easy way to say no, and a no still tells you something.
A last check before you send
Read the subject line the way the recipient will, in a crowded inbox, and ask four questions:
- Would they know why you're writing to them in particular?
- Does the start of the line make sense on its own if their phone cuts off the rest?
- Does the email keep the promise the subject makes?
- Would you be comfortable if they read it out to their team?
If any answer is no, rewrite it before you worry about testing it.
How to test without fooling yourself
Testing subject lines by reply rate is honest, and slow. Replies are rare, so small differences hide in the noise.
Here's the arithmetic. To tell a 3 percent reply rate from a 4 percent one with the usual standards (95 percent confidence, 80 percent power), you need roughly 5,300 sends per version. To tell 3 percent from 3.5 percent you need about four times that. Most sales teams don't send that much to one kind of prospect in a month.
So, practically:
- Change one thing at a time: length, or specificity, or personalization. Not all three.
- Send both versions to the same kind of list, at the same kind of time, so the list and the timing aren't doing the work.
- Count replies, and read them. A subject line that gets more replies saying "not interested" hasn't won.
- Give it weeks, not days, and be suspicious of any result from a few hundred sends.
What matters more than the words
Sender reputation. If your domain lands in spam, no subject line helps. SPF, DKIM and DMARC set up properly, and a clean list, come first.
Recipient relevance. A good subject line to the wrong person at the wrong company loses to a plain one sent to the right person.
Speed on warm threads. Once someone replies, the subject line has done its job. What matters then is how fast you answer, covered in the follow-up timing piece.
Using Outsolvi for this
Outsolvi grades every open as a person or a machine, records clicks, and labels replies Positive, Negative, Neutral or Out of office, so you can compare the emails you sent with each version by what people actually did. The 14-day trial needs no card. The open rate accuracy piece explains why the raw open count can't settle a subject-line test on its own.