Between November 2024 and May 2025, we had visibility into 50,000 B2B cold emails sent across a mix of SaaS, agency, and professional services teams. Same deliverability setup, varied personas, genuinely cold sends. We looked at subject lines specifically because the conventional wisdom hasn't kept up with what the inbox has become. Here is what the data actually shows.
What predicted higher open rates
- Length: 3 to 5 words. This is the single strongest pattern in the data. Two-word subjects underperformed slightly. Six or more words dropped off fast.
- Lowercase first letter. "quick question about {company}" beat "Quick Question About {Company}" by 11% open rate on average. It reads more personal.
- No punctuation. Question marks and exclamation points both hurt. A period at the end was neutral.
- Specific noun, not a generic promise. "pricing for acme" outperformed "quick proposal" by a wide margin.
- First-name personalization in the subject. "sarah — quick one" beat non-personalized subjects by about 14%, but only on first sends. Follow-ups lost the lift.
What hurt open rates
- "Quick question" on its own (not specific). Once one of the highest performers. Now a spam signal.
- Numbers in the subject. "3 ways to…" and "5 tips for…" both underperformed. These read as newsletter subjects, not personal ones.
- Emojis. Every test we ran showed emojis hurting opens for B2B cold. Consumer audiences may differ.
- All caps words. Even one word in all caps dropped open rate sharply.
- Re: prefixes on first sends. Tried. Hated. Don't.
What surprised us
Two things didn't match our expectations:
Urgency language was neutral, not negative. We expected "today", "this week", and "quickly" to tank open rates. They didn't. Buyers seem to have stopped reacting to these words either way.
Company name in the subject helped more than first name. "pricing for {company}" outperformed "sarah — pricing question" by about 7%. The company name signals relevance; the first name signals effort. Both work, but the company name wins slightly.
Open rate is not reply rate
The most important finding from the dataset is that subject lines optimized for opens don't always optimize for replies. Some of the highest-open subjects produced the lowest reply rates — they were curiosity-bait that didn't match the body. The subjects that actually correlated with meetings booked were the honest ones: specific, descriptive, not trying to trick the open.
"A subject that gets opened and then disappoints is worse than one that doesn't get opened. Open rates alone are a vanity metric in 2025."
The format that wins in 2025
Based on the data, the template that performed best across segments was:
{lowercase fact} for {company}
Examples from the winning pool:
- "pricing question for acme"
- "series b timing for acme"
- "hiring signal at acme"
- "mutual with jamie chen"
Short. Specific. Honest. No tricks.
What to do this week
Pull your last 500 cold sends. Group them by subject line pattern and compare open and reply rates. You'll probably find that 80% of your opens come from 20% of your subject patterns. Kill the long tail. Write new variants in the winning shape. Rerun the comparison in 30 days.