Cold email
Cold email subject lines: write and test
How to write honest subject lines, why open rate is a weak judge, and how to compare two lines by replies with a simple margin check.
Article
A subject line has one job: get the right person to open the message and know what it is about. Tricks that lift opens at the cost of trust tend to cost replies later.
This post covers what mailbox providers say about subject lines, six patterns you can adapt, and a test design that judges a subject by replies rather than opens, with a worked margin check on a fictional test.
What the providers say
Google asks that the subject, headers, display name and other message elements accurately represent the sender and content, and gives one example: do not start a subject with Re: or Fwd: unless the message is a real reply or forward. Yahoo tells senders not to use deceptive subject lines or misleading header information, and points to the CAN-SPAM Act.
The practical rule follows from this: the subject should describe what is inside. If you cannot say what the email is about in a few plain words, the email is not ready.
Six patterns to adapt
These examples are for a fictional scheduling company, Harborview Scheduling, writing to clinics. They are patterns to adapt to your own list, not lines proven to work. Word counts are computed from the text.
| Subject line | Words | Why it can work |
|---|---|---|
| Question about your onboarding calls | 5 | Names the topic, reads like a colleague wrote it |
| Harborview and Delmar Clinics | 4 | Names a real connection between two companies |
| Quick idea for reducing no-shows | 5 | States the outcome the message is about |
| Who handles patient scheduling? | 4 | A plain routing question |
| Your July webinar on intake forms | 6 | Points to something the person actually did |
| Following up on my note from Tuesday | 7 | Honest about being a follow-up, with no fake reply marker |
Why open rate is a weak judge
Open rate looks like the obvious score for a subject line. It is a poor one. Google states that it does not track open rates, cannot verify the accuracy of open rates reported by third parties and that a low open rate is not necessarily an accurate sign of a delivery problem. Open tracking relies on a hidden image that many mail programs load automatically or block, so the number moves for reasons unrelated to the subject.
Judge a subject by what happens next. Replies per delivered email is the cleanest measure you own, because a reply is an action a person took. The formula is replies divided by delivered, for each subject variant separately.
A test you can trust, with a margin check
Split one list at random into two equal groups, send the same body with different subjects at the same hours and count replies. In the fictional example, subject A had 12 replies from 300 delivered (4%) and subject B had 18 from 300 (6%).
B leads by 2 percentage points. To see whether that gap is more than luck, estimate the standard error of the difference: the square root of the sum of p(1 - p) / n for each group. Here it is about 1.8 percentage points. A gap smaller than about two standard errors is not strong evidence, and this one is 1.1 standard errors. So keep both variants running, collect more sends and do not declare a winner yet.
| Variant | Delivered | Replies | Reply rate |
|---|---|---|---|
| A | 300 | 12 | 4% |
| B | 300 | 18 | 6% |
| Difference (B minus A) | 2 percentage points, about 1.1 standard errors |
Rules of the road
Test one thing at a time, keep lines honest and write the subject last, after you know what the email says.
- Change only the subject between variants, never the body in the same test.
- Send both variants at the same hours so timing does not decide the result.
- Judge by replies per delivered email, not by opens.
- Do not fake a reply or forward marker.
- Write down each result so the next campaign starts from evidence.
Questions
- How long should a cold email subject line be?
- Long enough to say what the email is about and short enough to read on a phone. Test lengths on your own list and judge by replies.
- Can I use the recipient's name or company in the subject?
- Yes, when it is accurate and natural. It should point to something real about them, not just fill a merge field.
Keep reading
Cold email4 min read
Cold email templates and their logic
Five cold email templates for a sequence, each with the job it does, the size of its ask and its word count, ready to adapt to your own offer.
Cold email5 min read
Follow-up sequences that earn replies
Plan the touches, the gaps between them and the job of each message, then measure each step by replies per delivered email.
Cold email guide4 min read
Cold outreach 101: a guide for first sends
What cold outreach is, how a send turns into a meeting, which numbers to track, and what to set up before the first message leaves.
Put it to work
Pages in the product and the tools that go with this guide.
Sources
- Email sender guidelines, Gmail Help (Google): https://support.google.com/a/answer/81126, read 2026-10-19
- Sender best practices, Yahoo Sender Hub: https://senders.yahooinc.com/best-practices/, read 2026-10-19
- RFC 8058: Signaling One-Click Functionality for List Email Headers: https://www.rfc-editor.org/rfc/rfc8058, read 2026-10-19
Tell us what you sell and who buys it
What you sell
Who buys it
The button opens a draft in your own mail app. Nothing is sent from this page.
- Software companies
- Agencies
- Service businesses