Cold email subject lines that get replies

The subject line decides whether the email is read, and almost everything written about it is advice to be clever. For cold outreach the opposite is true: the goal is to look like an ordinary email from a person who has a specific reason to write.

What works

Short

Two to five words. Mobile clients cut off around 35 characters, and a long subject reads as marketing before it is even opened.

Specific to them

A subject that could be sent to ten thousand people looks like it was. Something that could only be sent to this person does not.

Weak:   Improve your sales process
Better: Your careers page / 3 SDR roles

Lowercase, unpolished

"quick question" outperforms "A Quick Question For You" because the first looks typed and the second looks composed. Title case is a marketing tell.

A real question

Questions invite an answer. "worth a conversation?" works because it is answerable in one word.

What to avoid

• Spam-trigger words: free, guarantee, act now, limited time, risk-free, exclusive offer.
• ALL CAPS and multiple exclamation marks.
• Emoji in cold B2B outreach. They read as bulk and some clients render them badly.
• Fake familiarity: "Re:" or "Fwd:" on a first email is a deception the reader notices immediately.
• Fake urgency: "URGENT" on an email that is not urgent costs you the reader permanently.
• The merge field that did not merge. "Hi {{first_name}}" in a subject is a familiar sight and it ends the conversation.

Patterns worth trying

Pattern	Example
Their company plus a specific thing	Acme / onboarding flow
A short question	worth a look?
An observation	noticed you are hiring SDRs
A referral, if genuine	Sarah suggested I reach out
Plain and direct	question about your pricing page

The referral pattern only works if it is true. If it is not, do not use it.

Testing properly

Most subject line "testing" is someone changing the subject and deciding from a handful of sends which one is better. That is noise, not a result.

To learn anything you need enough sends per variant that the difference is unlikely to be chance — around 30 sends per variant as a bare minimum, and more if the gap between them is small. Two variants, not five, so each one gets enough volume. And measure reply rate, not open rate: opens are unreliable and a subject that is opened but not answered has not helped you.

How RichLead handles this

You can define two subject styles and have them assigned randomly across your outreach. A contact keeps whichever variant it was given, so a follow-up does not switch style mid-conversation. Results are compared on reply rate, and a winner is only declared when the difference passes a statistical significance test rather than whichever number happens to be ahead.

Subjects can also be written per email by the model rather than from a fixed template, with the spam-trigger patterns above excluded.