TL;DR: A cold email subject line has one job: get a busy owner to open a message from a stranger. Keep it under 33 characters so it shows in full on a phone, name one specific detail, and skip the tricks. Replies, not opens, now tell you whether a subject line worked, so track replies from the first send.
Short, specific, honest subject lines get opened. Judge them by replies, because opens are no longer a clean signal.
Brafton and Lemlist both say keep the subject short, but they give different limits, so I checked the data for the rest. Brafton's rule is under 60 characters and 10 words (Brafton).
How short should the subject line be?
Keep the subject line under 33 characters. Mobile inboxes show only the first 33 to 50 characters, and a 2026 device test found 33 characters visible on every device it tested (WordCountChecker). Desktop shows far more, so a line that looks fine on your laptop can get cut mid-word on a phone.
If the phone cuts your subject mid-word, the reader never sees your point. I call this The Phone Cut-Off Test: read your line on a phone before you send it.
Twilio's guide puts the range in plain words: "Mobile inboxes only display the first 33 to 50 characters of a subject line while desktop inboxes display up to 70 characters in length." That is Julie Griffin writing for Twilio in January 2025 (Twilio). Twilio ran no test of its own, so read the range as a guide, not a law.
Here is where the advice gets messy. Belkins' 2024 report says the best subject lines run three, seven or eight words, with open rates up to 33%. Its 2025 study says two to four words won, at 46% open. The same company disagrees with itself across two editions.
So I would not chase a magic word count. Think of a subject like a shop sign: if the phone cuts it off after six letters, the customer walks past before reading the point. Which number do you trust? Test two lengths on your own list, to the same audience, on the same day.
These are my own drafts for a web agency, not tested lines:
- Your homepage on a phone
- Your booking page on mobile
- Quick question about your menu page
Does a first name still help?
Personalization lifted open rates in the data I found. Belkins' 2025 study of 2024 emails put personalized subject lines at a 46% open rate, against 35% for generic ones (Belkins). Its numbers do not separate a first name from a specific detail, so I cannot tell you which one did the work.
Put one specific detail in the subject, because a name alone says little about the business.
The Belkins report, written by Sophie Kompaniiets, puts it this way: "Emails with personalized subject lines boast a 46% open rate, compared to a mere 35% without." The same data shows replies rising from 3% to 7% for personalized lines (Belkins). I would watch that reply number first.
For a web agency, the detail has to be real and visible from outside: a slow homepage, a contact form that breaks on phones, or an old copyright year. LeadsAgent records that kind of evidence for each business, such as no website or an outdated site, so the detail you name is real, not a guess. Check the page once more before you send.
If you need that list first, how to get web design leads walks through the signals agencies use to find it.
Should the subject ask a question?
Maybe, but a question is not a shortcut. Belkins reports that question subject lines hit a 46% open rate (Belkins). That comparison is correlational, so it does not show the question caused the opens. Two controlled tests I found favored a plain statement, though both are old and aimed at non-service audiences.
Questions get opens in the data, but the controlled tests I found favored a plain statement.
NextAfter ran a nonprofit donor test in 2018 with 79,131 emails. A "Short and controversial" statement opened at 12.1%, and a generic question opened at 10.6% (NextAfter). In 2012, MarketingExperiments found an intrigue statement earned 47.9% more clicks on its call to action than a question, at over 99% confidence (MarketingExperiments).
Freelancers see the same pattern. Klento's author reviewed 2,000 cold emails and listed question-format subject lines among the worst performers, but gave no counts, so treat it as a lead (Klento). Lemlist says a question will not guarantee success, and that stacking question marks hurts opens (Lemlist).
The lesson sits in the body too. Reply.io's undated research says emails with no questions get 11% more replies, which is about body copy, not the subject (Reply.io). So when does a question fit? My rule: ask one only when the reader already asks it, and answer it in the first line.
Are interesting subject lines for emails worth the risk?
Curiosity can open a door, but a teaser that hides everything reads like clickbait. A freelancer's review of 2,000 cold emails lists curiosity-gap hooks among its worst performers, so a vague tease is a poor bet. Use interest only when the subject tells the reader what the email is about.
An interesting subject earns the open only when the reader knows what they are opening.
Klento's author put it bluntly: curiosity-gap hooks were among the worst performers (Klento). The author gives no counts, so I treat that as a warning, not a rule.
Google's sender rules add a second risk. They ask for content that is not misleading, so a teaser the email does not pay off can hurt more than it helps (Google).
Here is a pair I drafted to show the gap. These are my own lines, not tested ones:
- Teaser: "The thing I noticed on your site"
- Clear: "Your contact form on mobile"
The teaser makes the reader guess. The clear line tells them the topic before they tap.
What makes effective email subject lines for sales?
The strongest sales result I found came from a named colleague. Prospectio, a vendor, tested 7,340 prospects in 2025. It found replies rose from 27.0% to 35.1% when the email named a shared colleague (Prospectio). The test does not name the sector, so I treat it as a strong lead, not proof.
Naming someone the prospect knows lifted replies in the one large test I found. I call this The Shared Colleague Move, and it only works if the name is real.
Paul Cassidy, Prospectio's founder, sums up the result: "Naming a shared colleague lifted replies from 27.0 to 35.1 percent and meetings from 3.5 to 5.1 percent." His test also found that "Function matters more than seniority." (Prospectio). So pick a colleague whose job matches the prospect's.
Only name someone you have confirmed. A made-up connection misleads the reader, and Google's sender rules ask for content that is not misleading (Google).
Every subject line needs an inbox behind it, and LeadsAgent builds that part. It visits each business website, including its contact, about and team pages, and collects the public emails it finds. You then review the list and check the names yourself before you write.
If you want full emails to adapt, the cold outreach scripts for web design clients include six, from a cold email to a voicemail.
What should a follow-up subject do?
Stay in the same thread when you follow up, and never fake a Re: to get opened. Gmail's policy allows a Re: or Fwd: prefix only on a real reply or forward. Follow-ups carry much of the load, because Belkins' 2026 study of 2025 campaigns found they produced 58.6% of replies (Belkins).
A fake Re: line breaks Gmail's rule, so send one only on a real reply.
Gmail's help page is blunt: "don’t send messages with subject lines starting with Re: or Fwd: unless the messages are actual replies or forwards" (Gmail Help). Belkins gives the positive version. Its managing director, Alex Sribnyi, says to stay on the same email thread when you follow up (Belkins 2024 report)
The first email does a different job. Belkins' 2026 study put the per-step reply rate for first emails at 0.59%. I found no study that tests whether a Re: prefix changes replies, so Google's rule is the firmest ground I have.
For the messages themselves, the followup email guide covers when to send each one, and when to stop.
Think of a follow-up as a second knock at the same door, not a new house.
Is open rate still the number to track?
Not on its own. Belkins' 2026 study of 2025 campaigns did not track opens at all, and Twilio says opens are no longer a reliable signal (Twilio). Replies still mean something. The same Belkins study put the average reply at 0.45% of emails sent, so judge a subject line by the replies it brings (Belkins).
Open rate no longer tells you whether a subject line worked, so measure replies.
Opens are like a hand raised in a crowded room, and nobody can see whose hand it was.
Mailchimp's help center says clicks and purchases are stronger signs of engagement than opens, and that Apple's privacy feature does not affect them (Mailchimp). If your email has a link, track clicks next to replies. Freelancer numbers look different: Klento's own 2,000 emails averaged an 8.3% reply rate (Klento). Those numbers come from different senders, so use them as a range, not a benchmark.
Before you track replies on two subject lines, check that both lists can receive mail. LeadsAgent checks each email's format, flags disposable domains, reads the domain's mail server, SPF and DMARC records, and marks each row with is_deliverable. You can filter for verified emails only. It does not check the mailbox itself yet, so an address can pass the domain checks and still fail.
What to do instead: pull a list of verified public emails, send one subject per segment, and count replies until each version has enough to compare. LeadsAgent builds that list from public business websites, so you start from addresses the businesses publish.
Frequently asked questions
Does capitalization matter?
Belkins' 2025 study gave ALL CAPS subject lines a 30% open rate, against 25% for first-letter-only capitals (Belkins). The same report's Reply.io data gave uppercase subjects a 5.1% bounce rate, against 2.2% for lowercase (Belkins report). I would skip the shouting, because a bounce costs more than an extra open gains.
Do spam words sink a subject line?
I found no published test that isolates one word for a low-volume sender. An InboxEagle study of 774,828 emails found "free" had a 24.28% spam rate, against 22.56% with no trigger word. That sample is mostly bulk mail, not small outreach (InboxEagle). Google's sender requirements focus on authentication, a spam rate below 0.3%, and non-misleading content (Google).
Can a subject line mention an offer?
Yes, if the offer is true and the email delivers it. Google's sender rules ask for content that is not misleading. I found no measured test that compares offer lines with other lines, so judge them by replies.
Is there a cold call email sample?
Not with measured results. The pages I checked gave fragments, not full emails, and I found no reply data for cold call samples. The cold outreach scripts post above has six complete scripts.
Want the list before you run the test? Try LeadsAgent and see which businesses it finds on public websites.
