- Two to four words, specific, zero hype — that's the winning pattern.
- The subject line's only job is to not look like marketing.
- Test subject lines in batches of 100+, and judge by replies, not opens.
Your prospect triages email the way you do: thumb on phone, half a second per message, hunting for a reason to archive. The subject line's job in that half-second is surprisingly modest — it just has to not look like marketing. Everything that smells of a campaign gets archived on sight; everything that looks like a colleague's email gets a chance.
What actually works
The consistent winners across millions of sends are short, lowercase-ish, and specific. "quick question about your SDR hiring" beats "Revolutionize Your Sales Development Process 🚀" every day of the week. Internal-style phrasing — the way someone inside their company would write — is the strongest single pattern: "pipeline coverage for Q4", "re: the Austin office", "leads for the new AE team".
What kills opens (and lands you in spam)
- ALL CAPS, exclamation marks, and emoji — instant campaign signals
- Merge-field theater: "John, a special offer for Acme!"
- Vague curiosity bait: "You won't believe this..." — insults the reader
- Spam-trigger vocabulary: free, guarantee, limited time, act now
- Anything longer than ~6 words — it gets truncated on mobile anyway
The subject line can't save a bad list
A hard truth: subject lines move open rates by a few points; list quality and sender reputation move them by tens of points. If your open rates are under 40%, your problem is almost never the subject line — it's deliverability (authenticate your domain, warm it up, verify every address) or targeting. Fix those first, then optimize words.
How to test properly
Split-test one variable at a time on batches of at least 100 sends per variant, and judge by replies and meetings — not opens, which are unreliable since mail-privacy changes. A subject line that wins opens but loses replies is optimizing for the wrong click. Platforms like BixJet track reply outcomes per variant so the winner is decided by pipeline, not vanity metrics.