The complete cold email guide, from infrastructure to booked calls

I've sent more than 20 million cold emails and generated over $45 million in pipeline for clients doing it. This is the cold email guide I'd hand someone starting from scratch: the whole system on one page, in the order you actually have to build it, with the numbers each stage is supposed to produce.

Six stages. Infrastructure, lead list, offer, copy, follow-up, and everything that happens after the reply. Each one gets the compressed version here and a full breakdown of its own if you want to go deeper. The video below is the four-hour version. This page is the map.

Why the old playbook stopped working

Around 2020 this was easy. A couple thousand emails a month, 10 to 15% reply rates, meetings close to autopilot. Then it saturated, and the only answer anyone had was sending more than the next guy. All that extra volume is exactly what made Google and Microsoft crack down, and for a while only businesses that could afford a serious setup kept sending at all.

Then the opposite happened. So many sending tools launched that the only thing left to compete on was price, and what used to cost $1,000 to $2,000 a month now runs $200 to $500. Sending at scale got cheap. Landing in the primary inbox did not. That gap is the entire game now, and the six stages below are how you close it.

Stage 1: infrastructure

You cannot fire 10,000 emails a day out of your main domain. You spread the sending across lookalike domains and dozens or hundreds of sending accounts, all managed from one sequencer so replies land in a single inbox.

The math, for a 10,000-a-day setup: roughly 60% Microsoft and 40% Google. A Google account sends 10 to 15 emails a day and needs one domain per three accounts. A Microsoft account sends 5 a day and runs 99 accounts off a single domain. Stack that up and you need about 113 domains. Extension matters less than people think: .com is best for deliverability, .co second, .info third, but .info runs about $2.57 a year against $10 to $12 for a .com, and at this scale the extra domains you can afford outweigh the difference.

Resellers handle SPF, DKIM and DMARC for you now, which used to be the hardest part. What they cannot skip is warmup: two weeks of the accounts mailing each other before you send a single prospect, so the mailbox providers treat them as real. Read Google's sender guidelines once before you build anything, because bulk senders live or die by the bounce and complaint thresholds in that document.

Starting from zero, ignore the 10,000-a-day math entirely and build for 10,000 to 20,000 sends a month. That is 20 to 25 Google domains for about 50 accounts, plus a single Outlook domain running 99 accounts at 5 sends a day. The full domain, inbox and warmup walkthrough is in the setup guide, and what the whole build costs at scale is in the infrastructure breakdown.

Stage 2: the lead list

Most people open Apollo, which holds over 275 million contacts and 73 million companies, add a few filters based on who they think their ideal customer is, build one enormous list, and send. The filters are usually copied from inbound behaviour, and inbound and outbound are two different populations. Inbound already knows you and knows they have a problem. Outbound may know neither.

Two things fix the list. First, build it out of keywords rather than industry filters, which are broad and overlap in ways that quietly wreck a campaign. Add employee count on every single list, cut the one-to-ten-person companies that cannot afford you and the enterprise accounts you cannot break into, and use time in current role as a proxy for an established business. Ignore the built-in buying-intent and email-open filters entirely; they are not accurate enough to act on.

Second, do not export from Apollo. Scrape the search and pay per lead: 15,000 leads runs about $29, against hundreds of dollars in credits for the same list. Then clean it in Clay: normalize the company names, waterfall the missing emails through several providers, and verify everything before you spend a cent on personalization. Even a talented Apollo user is working with about 70% accuracy without that step.

Then segment and test. Every segment needs at least 1,000 sends before you are allowed an opinion about it. One client's launch ran five segments across 14 sub-segments: 14,000 prospects, two emails each, 28,000 sends over 10 sending days, so 2,800 a day. Same offer, same copy, only the segment and the data source changed. Three creator segments came back at 12 to 18% interested. Two agency segments, built from that client's best case studies, came back at 1%. Nothing about the copy caused that. The full filter-by-filter breakdown lives in the Apollo walkthrough.

Stage 3: the offer

This is the 80/20 of the entire system, and it is where most campaigns are lost before a single email goes out. You are a stranger asking someone to commit to a $10,000 retainer off four sentences. Nobody does that. So the thing you put in the email is not your service, it is a front-end offer: a specific, tangible thing they get for replying, free or close to it, that costs them minutes rather than hours.

The formula I use is the most value you can deliver for the least work on your end. A free deliverable beats a description of a deliverable every time. And the default ask in almost every industry, the free audit, is the weakest thing you can send: an audit is a diagnosis, a diagnosis from a stranger is just an opinion, and it leaves them with homework. Nobody wants an audit from someone they have never heard of.

Test three or four shapes at once. Something delivered (five ad creatives, one edited video, a rebuilt listing), something diagnostic but reframed around what they are losing rather than what they might gain, and something asset-based they keep either way. Then judge them by booked calls, never by how good the reply rate feels. Fifteen worked examples across fifteen industries are in the front-end offer breakdown.

Stage 4: the copy

Once the offer is right, the copy does not have to sell anything. Its only job is presenting that offer in the most casual, frictionless way available. People are so conditioned to marketing email that the moment a message smells like persuasion, they stop reading, which is why long personalized essays lose to four lines that get to the point.

Subject lines: two or three words, lowercase, curiosity-inducing, and they should read like internal communication. "capital stack question" and "Instagram question" both work because they hint at the topic without naming the offer. Name the offer in the subject line and there is no reason left to open the email.

Then the body, which is shorter than anyone expects:

{first name}, if we could create five ad creatives for you completely free and prove which ones convert before you commit to anything, would that be worth sharing more?

No greeting. The name sits at the front like a text message, the offer is the whole message, and the call to action is deliberately low friction. "Worth sharing more" outperforms "worth a chat" because it can resolve into an email, a video or a call, and the reader is not agreeing to a calendar invite yet. A PS line carrying one piece of social proof is worth testing, though it sometimes triggers sales resistance, so run it both ways. Keep the proof in plain money terms; nobody outside your industry knows what your acronyms mean.

Build four variations from two subject lines and two bodies, run them through spintax so thousands of sends are not identical, and change exactly one variable per test. Twenty-seven differences between two emails tells you nothing. The full copy breakdown has the frameworks and the before-and-after rewrites.

Stage 5: the follow-up

Most of the booked calls do not come from email one. People are busy, inboxes are flooded, and plenty of prospects see email two first without ever opening email one. That single fact sets the rule: every follow-up has to stand on its own, with the offer restated, so nobody is forced to scroll up a thread to understand what you want.

The move that works is lowering the friction each step instead of raising the pressure. If email one offers the deliverable, email two offers a two-minute video explaining how it works. Same value, easier yes. Two or three days between them. And if your first email is a single line, the follow-up is where the things that did not fit go: the objection you know is coming, the reason you are doing the work for free, the risk reversal.

Once someone replies interested, the follow-up turns multi-channel. We aim to respond within five minutes, which is why every interested reply fires a webhook that enriches the prospect's mobile number and pushes it to Slack as a clickable link. Then a warm call, then a LinkedIn connection request with no message attached, because that gets accepted more often and puts your content in front of them for the next two weeks. Email, phone, LinkedIn, in that order, is covered in the follow-up system breakdown.

Stage 6: everything after the reply

This is the stage nobody teaches and the one that decides whether any of the previous five were worth doing. Interested replies are not booked calls, and booked calls are not calls that happen.

Start by categorizing the inbox with AI: no, interested, meeting request, more information, out of office. Then push toward a meeting as fast as you can while actually answering the objection, rather than running a long email conversation that goes nowhere. Suggest two specific times and book them yourself when they pick one.

Qualify before the call, not on it. Two or three questions on the booking form is usually right: website, revenue, deal size. Ours also states plainly who the system works for, including a market of at least 50,000 prospects, because saying so filters out the people it would fail. Treat qualification as a lever, not a setting. Too tight and you cut good calls; too loose and you burn the calendar. Budget, authority, need and timing is the frame. The exact questions, the confirmation page and the nurture sequence behind them are in the qualification breakdown.

Then the pre-call sequence, which is the highest-return thing on this page relative to effort. A confirmation page with a video under two minutes explaining why the page exists, then short Q&A videos answering the objections you already hear on every call: how the process works, who is on the team, how long results take, what the investment looks like. Behind that, a nurture sequence of case studies and problem framing, plus reminders at 24 hours, 4 hours and 10 minutes by both email and text. People do not no-show because they forgot. They no-show because they do not believe you yet, and reminder emails alone have never fixed a belief problem.

The numbers to judge it by

Every stage above has a number attached. If a campaign is missing its target, this table tells you which stage broke rather than leaving you to guess.

Building it in order

The sequence matters, because each stage is useless without the one before it:

1. Order domains and inboxes, and start the two-week warmup today. It is the only step you cannot compress.
2. While it warms, write the front-end offer and test three or four versions of it on paper.
3. Build one list, segmented, with at least 1,000 prospects per segment. Verify every email.
4. Write one email and one follow-up per offer. Four variations, one variable apart.
5. Send for two weeks, then cut the losing segments and double the winners.
6. Build the qualification form and pre-call sequence before the first call happens, not after the first no-show.

One honest caveat before you start. This works when there are enough reachable buyers to find the winning offer, and it does not work when there are not. If your entire market is a few thousand companies, you will run out of people before the data tells you anything, and no amount of infrastructure fixes that. I'd rather you know that now than after you have bought 113 domains.

tl;dr: cold email is six stages, not one, and the offer is the 80/20 of all of them. Get the list precise and the offer real, and the email itself can stay plain.

FAQ

If you'd rather have all six stages built and run for you, offer, lists, copy, infrastructure, and a setter working the replies, that's what we do: 3,000+ sales calls booked for clients in the past year. Or keep learning free. Every training, grouped by topic, sits on the trainings hub, and the full four-hour version of this one is on YouTube.

PS - the campaign in the stat plate above used two emails per prospect and no clever personalization. The frame did the work: they were told a buyer was already looking.