I Analyzed 5,000 B2B Cold Emails to Find What Actually Gets Replies — Here's the Data

I Analyzed 5,000 B2B Cold Emails to Find What Actually Gets Replies — Here's the Data

Last year, I hit a wall. I was running outbound for a B2B SaaS product I'd been building, sending about 50 personalized emails a day, and getting maybe one or two replies a week. The classic "spray and pray" approach wasn't working, but all the advice I found online was either generic ("personalize better!") or straight-up contradictory ("use humor!" vs "be professional!").

So I decided to stop guessing and start measuring. I built a system to scrape, categorize, and analyze cold emails — both my own and ones I was looped into through various B2B networks — to figure out what actually separated the replies from the black holes.

Here's what the data says after processing over 5,000 outreach emails.

The Setup

I compiled a dataset from three sources:

• My own outbound campaigns (approx. 1,200 emails sent over 8 months targeting SaaS founders and marketers)
• Emails shared publicly in B2B founder communities, sales forums, and case studies (about 2,300)
• Emails I received as a founder running a product (roughly 1,500 from various vendors, agencies, and tool makers)

For each email, I tracked: subject line structure, first sentence approach, personalization depth, length, call to action, timing, industry, and — crucially — whether it got any reply at all.

Surprising Finding #1: Length Matters Way Less Than You Think

Conventional wisdom says keep cold emails under 100 words. The data mostly supports this — emails under 120 words had a 5.8% reply rate vs 3.1% for longer emails. But here's the twist: the really short emails (under 50 words) performed worse (3.9%) than the 80-120 word sweet spot (6.4%).

The issue with ultra-short emails is they feel templated. "Hey, saw your company, love what you're doing. Would you be open to a 15-min call?" — I received this exact email from 12 different vendors. None of them got a reply from me.

The winning approach was 90-110 words with specific personalization. Not "I see you work at [Company]" but "I noticed [Company] just launched [Feature X] — I've been following how you're approaching [Problem Y] because we're solving something related."

Surprising Finding #2: Timing Is a Real Effect, But Not the One You've Heard

You've probably read that Tuesday at 10 AM is the best time to send. My data says this is wrong — at least for B2B SaaS founders. The best reply rates in my dataset came from:

1. Thursday between 2-4 PM (7.2% reply rate) — people are wrapping up the week, less protective of their morning focus time
2. Wednesday around 11 AM (6.1%) — mid-morning, post-rush
3. Sunday evening 6-8 PM (5.8%) — surprisingly strong for founders, who often catch up on emails over the weekend

Monday mornings and Friday afternoons were dead zones. Tuesday at 10 AM was middle of the pack (4.1%). Not terrible, but certainly not the magic slot everyone claims.

Surprising Finding #3: The Personalization That Matters Most

I categorized personalization into three tiers:

• Tier 1 (shallow): Using the recipient's name and company name
• Tier 2 (moderate): Mentioning a recent event, blog post, or product update from the recipient
• Tier 3 (deep): Referencing something specific about their work — a GitHub commit, a talk they gave, a specific problem their product solves differently

Tier 1 barely moved the needle (3.8% vs 3.2% for zero personalization). Tier 2 bumped it to 6.1%. Tier 3 hit 11.4% — nearly 3x the baseline.

But here's the catch: Tier 3 is expensive. Each deep-personalized email took me 5-10 minutes to research and write. At that rate, sending 30 emails a day is a full-time job. Most people simply can't scale it.

What I'd Do Differently

Looking back, there are two things I got wrong:

I optimized for the wrong metric. I was tracking emails sent per day instead of qualified conversations started. The emails that took me 30 seconds to write rarely led anywhere. The ones that took 5 minutes often led to real conversations. I should have been merciless with my time: if I couldn't find something real to personalize, I shouldn't have sent the email at all.

I undervalued follow-ups. The data showed that a three-email sequence (initial + two follow-ups spaced 4-5 days apart) had a cumulative reply rate of 12.3% — more than double the single-email rate. Most people give up after one email. Persistence, done respectfully, works.

The Problem with Doing This By Hand

The deeper I got into this, the clearer it became that the bottleneck isn't writing emails — it's research. Finding something genuinely personalized about a prospect (a recent blog post, a GitHub commit, a product launch) takes real digging. Doing that 30 times a day is exhausting.

I ended up building clienthunter.ai to automate the research part — it scours the web for signals about prospects: what they're building, what they've shipped recently, what problems they're publicly wrestling with. It turns the 5-minute research grind into 10 seconds, so I can focus on the writing and relationship part that actually needs a human touch.

The Big Question I'm Still Trying to Answer

The data is clear that deep personalization works. But there's a tension I haven't resolved: at what point does "personalized outreach" cross into "creepy"? I've had prospects tell me they found it impressive that I referenced their recent product launch — and others who seemed unsettled that I knew about it.

Where do you draw the line? How much research is too much before reaching out?