Learn / Infrastructure
How to automate cold email outreach
Most people send 10 to 20 cold emails a day by hand, chase the replies in their own inbox, and call that outreach. We send 10,000 a day and nobody touches an individual email. Learning how to automate cold email outreach is really four decisions: which platform you send from, how you structure the campaign, what you split test, and what happens automatically the second somebody replies.
This page is the written version of the training below, and it stays on the workflow. How many domains, how many inboxes, what each mailbox costs: that math already lives in the infrastructure post. What follows is the part you operate.
Pick a sequencer you can build on top of
You cannot run this out of your normal Google or Microsoft account. Volume from one mailbox gets flagged as spam within hours, but the bigger problem is that a normal mailbox has none of the plumbing: no rotation across accounts, no per-account caps, no reply routing, no way to push a lead into a follow-up sequence when they answer. Three platforms are real options.
| Platform | Who it fits | The tradeoff |
|---|---|---|
| Smartlead | Deep automation without applying for access | Shared reputation and a shared warmup pool, plus enough shipping velocity that uptime wobbles |
| Instantly | A first campaign: the interface carries nothing but what you need | Same shared reputation, and per-client plus per-lead pricing that stops being cheap once you stack accounts |
| EmailBison | Many client campaigns at once, and building on the API | Pricier, and you have to apply to become a user at all |
We ran Smartlead for about a year and a half and it was genuinely good. The line that eventually decided it for us was reputation. Smartlead and Instantly are mass-market tools, so your sending reputation is pooled with everyone else on the platform, warmup pool included. You inherit other people's sending behavior whether you like it or not. EmailBison isolates it, and that is why we moved. Emails land in the primary inbox more often, and better reply rates come out of that. The head-to-head on the first two is in the Smartlead vs Instantly comparison.
The other thing worth weighing is what you can build on top. We run our whole business off the API: campaign creation, reply routing into Slack, reporting, client dashboards. The friendlier tools are cheaper for one sender and charge per client and per lead, which is fine right up until you stack accounts.
Warm every inbox for two weeks, then leave it alone
Once the mailboxes are bought, the reseller handles SPF, DKIM, and DMARC and pushes the accounts into your sequencer as senders. Most start warmup automatically too. Ours run around 5 warmup emails a day on the Outlook accounts and closer to 15 a day on Google, and the honest instruction is to follow whatever your provider recommends instead of pushing the number up.
What is NOT flexible is the two weeks. Let warmup run for 14 days before a single campaign email leaves those accounts. Everything past this point is automation, and automation on half-warm infrastructure burns the domains you just paid for before you learn whether the offer works. Google's bulk sender guidelines are the rules you are warming into, and Microsoft has its own equivalent.
Clean the list before it ever touches the sequencer
Export the list from a platform like Apollo, then pull it into Clay to clean it up. Verify everything with MillionVerifier and keep only the contacts that come back deliverable, so you are not paying in bounce rate to learn an address was dead. One enrichment on top of that is usually worth it, and the useful one is the company's actual industry, because that is the field your copy references. Then export the CSV.
The upload is where people quietly create a problem for themselves six weeks later. Map every field deliberately and drop the ones you will never use. Name the list. Set existing leads to update rather than duplicate. And turn parallel sending off, so a lead can only sit in one campaign at a time. Two campaigns emailing the same person in the same week is how a good list turns into a complaint. Starting from zero on the whole motion, the wider map is the complete cold email guide.
Name and configure every campaign the same way
Every campaign gets the same name structure: company, industry, a descriptive filter, then the date. A lending campaign for companies with 10 to 50 employees launched on November 2nd is named exactly that. It sounds trivial. Once you are running dozens of campaigns across a dozen clients, the naming convention is the only thing between you and a reporting problem you cannot untangle.
The settings are identical on every campaign, which is the point. Subject line stays boring, and {{first_name}}plus something like "FYI" is tried and true. The body carries the name variable, your signature, and whatever personalization you chose. This plain is fine:
{{first_name}}, are you looking for new clients to provide loans to this quarter?
Sending window 8am to 5pm EST. Plain text, always. Do not track opens, at cold volume or anywhere else. Attach every sender email you set up, then run a preview before you launch. The preview is how we caught a signature rendering wrong on a handful of accounts, which nobody would have noticed until it had gone to a few thousand people.
Automation multiplies whatever you built. Point it at a dead offer and it multiplies that.
Split test angles, not adjectives
Build the split tests in before you launch, not after the campaign disappoints you. Three to four variants, and on a first campaign genuinely different ones: separate calls to action, separate versions of the offer, separate angles. Not word tweaks. You are trying to find which direction is alive, and a synonym cannot answer that. Hyper-optimization comes once one angle is clearly working.
Same logic on the steps inside the campaign. Two to three, no more. Email two should say something email one did not, usually with a lower-friction ask: if the first pushed for a call, the second can offer a short video instead. Draw it out further and at cold volume the extra steps buy unsubscribes rather than replies.
Subsequences are where the automation earns its money
Campaigns go live, replies land, and almost nobody books straight off the first one. That is what subsequences are for: an automated follow-up campaign a lead gets pushed into based on what they actually said, running on time delays so you are not tracking every person who showed interest once and went quiet.
The rule is one subsequence per response type. Somebody who asked for the free video does not belong in the same follow-up as somebody who asked for a meeting. If you handed over an asset, the follow-up's job is making sure they used it, not asking again for a call. Matching the sequence to the answer is the whole trick.
Tone matters more here than anywhere else in the system. No "just bumping this up", no "following up", no "how's it going". That is pestering, and it is what makes people mute the thread. Every touch should add something. The barebones meeting-request subsequence runs three:
- Checking back to see if they are still interested, booking link reshared. Better if it carries an asset or a short video that explains more.
- Making sure this does not slip through the cracks, offering time in the next few days. The natural place for a case study.
- "It looks like you may have gotten caught up." Then name the outcome: if building a system that generates new business opportunities is still a priority, let me know.
Setting it up is the easy part. You create it as a subsequence rather than a campaign, leads get pushed in automatically from your inbox based on the reply, and the sequence, schedule, and settings are configured exactly like a normal campaign. What to say across those touches, and how fast to move, is in the follow-up system breakdown.
The order is the part people get wrong
Doing this out of order is how people burn every domain they bought before sending anything worth measuring. Platform, warm infrastructure, verified list, a campaign structured the same way every time, split tests, then a subsequence behind each reply type. And the honest caveat: automation is a multiplier, and multipliers do not care what they multiply. A weak offer sent 10,000 times a day is just a faster way to learn the offer is weak.
tl;dr: pick a sequencer you can build on, warm every inbox for two weeks, verify and map the list once, configure every campaign identically, split test three to four real angles across two to three steps, and put a subsequence behind every reply type.
FAQ
What is the best platform to automate cold email outreach?
Smartlead, Instantly, and EmailBison are the three real options. Instantly is friendliest for a first campaign, Smartlead sits in the middle with deep automation, and EmailBison is the power-user pick because sending reputation is isolated instead of pooled with every other account. If you plan to build automations on top of the sequencer, pick on API depth.
Can you automate cold email from Gmail or Outlook directly?
No. A normal mailbox has no account rotation, no per-account sending caps, no reply routing, and no way to push a lead into a follow-up sequence when they answer. Volume from one mailbox also gets flagged as spam almost immediately. You need a sending platform, with your mailboxes attached to it as senders.
How long before an automated campaign can start sending?
Two weeks of warmup on every new inbox before a single campaign email leaves it. Most providers start warmup automatically once the accounts land in your sequencer, around 5 warmup emails a day on Outlook and closer to 15 on Google. Launching early on half-warm accounts burns the domains you just paid for.
How many split test variants should a cold email campaign have?
Three to four on a first campaign, and they need to be genuinely different: separate offers, separate calls to action, separate angles. Word-level tweaks cannot tell you which direction works, so they waste the volume. Once one angle clearly wins, that is the point to start testing smaller changes inside it.
How many follow-up emails should a cold email campaign have?
Two to three steps, no more. Email two should add something the first one did not say, usually with a lower-friction ask: if email one wanted a call, email two can offer a video or an asset instead. At cold volume, seven-step sequences buy unsubscribes and complaints rather than pipeline.
What is a subsequence in cold email?
An automated follow-up campaign a lead gets pushed into after they reply, based on what they said. One per response type: whoever asked for a meeting gets the meeting-request follow-up, whoever asked for a free asset gets the sequence that makes sure they used it. It is what stops warm replies going cold.
If you would rather have the whole system installed for you, platform through subsequences, that is what we do: 3,000+ sales calls booked for clients in the past year. Or keep learning free, every training grouped by topic on the trainings hub, and the full video for this one on YouTube.
PS - the first campaign you automate will be your worst one. Build it so the second one takes an hour instead of a week, and the compounding starts there.