Cold email has a reputation problem, and mostly it is deserved. The average founder's first outbound campaign is a template scraped from a blog post, blasted to a purchased list, opening with "I hope this email finds you well" and closing with a request for "15 minutes to pick your brain." It gets a two percent reply rate on a good day, most of those replies are unsubscribes, and the founder concludes that cold email does not work.
Cold email works. What does not work is treating it as a volume game when you have neither the list quality nor the domain reputation to survive one. For an early-stage company, outbound is a research exercise with an email attached — and when you approach it that way, the numbers change completely.
The Uncomfortable Math
Start by being honest about what you are trying to achieve. If you need ten conversations this month and you convert roughly one in ten replies into a call, you need a hundred replies. At a realistic five percent reply rate for well-targeted outbound, that is two thousand emails.
That number is why most founders reach for automation and buy a list. It is also why most founders fail. Two thousand poorly targeted emails will destroy your sending domain before they book a single meeting.
The alternative is to change the inputs rather than the volume. A tightly researched list of two hundred genuinely relevant prospects, emailed with specificity, can produce a twenty to thirty percent reply rate. That is sixty replies from two hundred emails instead of a hundred from two thousand — with a tenth of the work and none of the reputational damage.
Narrow beats broad, every time, at this stage.
Build a List You Could Defend
The single biggest predictor of outbound success is list quality, and list quality means one thing: could you explain, in a sentence, why this specific person has the problem you solve?
If the answer is "they work at a company in our target industry," that is not good enough. If the answer is "they posted last week about struggling to keep their onboarding docs current, which is precisely what we automate," you have a real prospect.
Good sources of that kind of signal include job postings that reveal a company is building the thing you replace, public changelogs or engineering blogs that indicate what a team is working on, conference talks and podcast appearances, questions asked in community forums, and reviews left on competitor tools. Each of these tells you something about a specific situation rather than a demographic bucket.
This research is slow. Twenty to thirty prospects an hour is a reasonable pace. It is also the work that makes everything downstream effective, and it cannot be delegated to a tool at this stage because the judgment is the value.
The Anatomy of an Email That Gets Read
Your email is competing with a hundred others, most of which look identical. Three things separate the ones that get replies.
It is obviously not a template. The first sentence must contain something that could only be written to this person. Not their first name and company — any mail merge does that. A reference to something they actually said, shipped, or published. This single sentence does more work than the rest of the email combined, because it answers the reader's first question: is this real?
It is short enough to read on a phone. Under a hundred and twenty words. No logos, no images, no formatted signature block with four social icons. It should look like an email a person typed, because that is what gets replies.
It asks for something small. "Would you be open to a 30-minute demo" asks a stranger for half an hour based on nothing. "Is this something your team deals with?" asks for a yes or no. Get the conversation started and earn the meeting in the reply, rather than demanding it up front.
Here is the shape that works:
Hi Sana — saw your post about the support queue doubling after the Series A. We built [product] because the same thing happened to us at 30 people and none of the tools we tried handled the triage part.
Is manual triage still eating your team's mornings, or have you solved it since?
— Alex
That is sixty words. It demonstrates research, states relevance without pitching, and asks a question a busy person can answer in one line.
Subject Lines Are Not Where the Creativity Goes
Founders spend disproportionate time on subject lines, usually making them worse. Clever subject lines look like marketing. Marketing looks like something to archive.
The best-performing cold subject lines are boring and specific: a few plain words that describe the topic. "Support triage," "your onboarding docs," "question about your API rollout." They read like internal mail, which is exactly the point.
Avoid anything that resembles a promotion — no capitalisation gimmicks, no emoji, no false urgency, no "quick question" (which every automated sequence in the world now uses and which readers have learned to ignore).
Follow Up Twice, With New Information
The majority of replies to cold outreach come from follow-ups, not the first email. But most follow-ups are worse than useless because they say nothing new: "Just bumping this to the top of your inbox" adds no value and signals that you have nothing else to offer.
Send two follow-ups, spaced four or five days apart, and make each one carry fresh information. A relevant customer result. A short answer to an objection you expect. A link to something genuinely useful whether or not they ever buy. Then stop.
The final email in a sequence should make it easy to close the loop: "I will assume this is not a priority right now and leave you be — happy to reconnect if that changes." This is polite, it respects the reader, and it produces a surprising number of "actually, wait" replies.
Three emails total. More than that and you are not persistent, you are a problem.
Protect Your Domain Like It Matters
It does matter. If your primary domain gets flagged as a spam source, your transactional email — password resets, receipts, invoices — starts landing in junk folders, and that damage is slow and painful to reverse.
Buy a separate domain for outbound. Set up SPF, DKIM, and DMARC properly before the first send. Warm it up gradually over two to three weeks rather than starting at full volume. Keep daily send volume modest — dozens, not hundreds, per mailbox. Remove anyone who does not engage, and honour opt-outs instantly and permanently.
Also, know the rules that apply to you. Depending on your recipients' jurisdiction, unsolicited commercial email carries real legal requirements around identification, opt-out mechanisms, and in some regions consent. Being compliant is not optional, and it mostly overlaps with being decent anyway.
Measure Replies, Not Opens
Open rates have become close to meaningless — privacy features in mail clients fire tracking pixels automatically, inflating the number for reasons unrelated to human attention. Optimising for opens optimises for noise.
Track reply rate, positive reply rate, and meetings booked. Those are the numbers that connect to revenue. If your reply rate is under five percent, the problem is almost always the list rather than the copy. If people reply but nothing converts to a call, the problem is your ask.
Outbound Is a Learning Channel First
The most valuable output of early cold email is not the meetings. It is the language. Every reply — including the rejections — teaches you how your buyers describe their problem, which objections come up first, and which framings land.
That intelligence flows directly into your landing page, your onboarding, and your positioning. Founders who do outbound personally, at least for the first few hundred emails, end up with sharper marketing than founders who never talk to strangers about their product.
Do it yourself. Keep the list small and the research deep. Write like a person. And stop after three.