What the data actually says
Most 'best subject lines' posts are opinion pieces disguised as data. Here's what we see across real B2B cold email campaigns in our system.
We analyzed subject line performance across campaigns with 500+ emails sent, measuring open rate and reply rate (not just opens — a subject line that gets opened but never replied to isn't converting).
Length: shorter wins
Subject lines with 4–7 words consistently outperform longer alternatives. The data:
1–3 words: high open rate but feels incomplete; often filtered as spam 4–7 words: the sweet spot — 15–20% higher open rates than 8+ words 8–12 words: acceptable but gets truncated on mobile 13+ words: actively harmful — truncated everywhere, feels promotional
The mechanism is simple: shorter subject lines display fully on mobile (where 60%+ of B2B email is first seen), and they feel personal. A 4-word subject line looks like it was written by a human, for one person. A 12-word subject line looks like marketing.
Specificity: the single biggest lever
Subject lines that reference something specific about the recipient — their company name, a recent event, a metric, a mutual connection — outperform generic subject lines by 28–42% on reply rate.
Examples of generic vs specific:
Generic: 'Quick question about your outbound strategy' Specific: 'Saw the SDR role posted last week'
Generic: 'Improving sales efficiency' Specific: 'Re: the Q2 headcount freeze'
Generic: '10x your pipeline' Specific: '3 accounts you're missing in automotive'
The specific versions work because they signal 'I actually know something about you.' The generic versions work only if the recipient happens to be in the market — which is random chance, not strategy.
What to avoid (the spam triggers)
Certain patterns reliably trigger spam filters or reduce open rates:
- ALL CAPS or Title Case On Every Word — both signal mass email. Use sentence case.
- Spam trigger words: 'free,' 'guaranteed,' 'exclusive offer,' 'act now,' 'limited time.' These have been spam-flagged since 2005 and still trigger filters.
- Re: or Fwd: on a first-touch email. Some tools add fake reply prefixes to boost open rates. It works short-term, destroys trust long-term, and increasingly triggers spam filters.
- Excessive punctuation: '!!!,' '???,' or emoji in B2B contexts.
- Asking a question that the recipient has no reason to care about: 'Want to 10× your pipeline?' No.
Four formulas that consistently work
These aren't templates — they're structural patterns. Fill them with something specific to the recipient:
1. The observation: '[Specific thing I noticed]' — e.g., 'Saw the Portland warehouse expansion' 2. The mutual context: '[Shared connection / event / group]' — e.g., 'From the Pavilion thread on SDR comp' 3. The metric: '[Relevant number] at [their company]' — e.g., '3 open AE roles at Acme' 4. The question: '[Specific question about their situation]' — e.g., 'Still evaluating after the Outreach renewal?'
All four share a common trait: they prove you did research. That's the whole game. A subject line that proves you know something about the recipient will outperform any 'best subject line template' regardless of format.
Frequently asked questions
What is the best cold email subject line length?
4–7 words. Subject lines in this range have 15–20% higher open rates than longer alternatives. They display fully on mobile (where 60%+ of B2B email is first seen) and feel personal rather than promotional.
Should I use the recipient's name in the subject line?
It depends on the context. A first name alone ('Ayush') can work if the rest of the email is genuinely personalized. But {{First_Name}} in the subject line of a clearly templated email signals 'mass outreach' and hurts more than it helps.
Do emojis work in cold email subject lines?
For B2B cold email to mid-market and enterprise: no. Emoji subject lines correlate with 10–15% lower open rates in our data. They signal consumer marketing, not professional outreach.
Should I A/B test subject lines?
Yes, but only if you're sending enough volume for the test to be statistically meaningful. Below 200 emails per variant, the data is noise. Above 500, patterns become reliable.