Most cold email campaigns fail before the first send. The decisions that determine success happen in the two weeks before you write a single line of copy.
Launching cold email correctly is a sequencing problem. Infrastructure first, then targeting, then messaging. Teams that reverse this order end up rewriting copy to fix what is actually a deliverability or ICP problem — and never figure out why it isn't working.
This is the sequence that holds.
Step 1: Define your ICP before anything else
The most expensive mistake in outbound is sending a good message to the wrong person. Infrastructure can be rebuilt. Reputation can recover. But the signal you get from a misaligned ICP leads you to wrong conclusions for months.
ICP definition is not a marketing exercise. It is a filtering exercise. You are not describing every company that could theoretically buy — you are identifying the profile where your offer lands with the least friction.
Start with your existing customers. What do the ones with the shortest sales cycles have in common? Industry, company size, tech stack, growth stage, headcount in a specific function — these are signals. If you don't have customers yet, work backwards from your offer: who feels the problem most acutely, has budget to solve it, and has a decision-maker reachable by email?
Practical filters to define before you build a list: industry vertical, company headcount range, geography, job titles you're targeting, and one or two signals of intent (recent funding, hiring in a specific role, new product launch).
Step 2: Define what you want recipients to do
Every sequence needs one CTA, not three. The most effective cold email CTAs are low-friction asks: a 20-minute call, a yes/no question, a single link to a case study. The moment you ask for too much in a first email, reply rates collapse.
Define the CTA before you write the email. The body of the email exists to earn the ask — you can't write it well if you don't know what you're building toward.
Step 3: Set up your sending infrastructure correctly
Your primary domain should never send cold email. Register secondary domains specifically for outbound — variations of your main domain work fine (.co, .io, a prefix or suffix). One domain supports two to three mailboxes. Each mailbox should send no more than 30–50 emails per day at steady state.
For each domain, configure three things before a single email leaves it:
- SPF — authorises which servers can send on behalf of your domain
- DKIM — cryptographically signs outgoing mail so receiving servers can verify authenticity
- DMARC — tells receiving servers what to do when SPF or DKIM fails, and gives you visibility into spoofing attempts
Skip any of these and deliverability degrades regardless of what you do downstream. Most domain registrars and sending tools have step-by-step setup guides — there's no excuse for skipping them.
For the sending tool itself, Google Workspace and Microsoft 365 (Outlook) are the two credible options for cold outbound. Outlook consistently achieves better inbox placement when sending to Gmail addresses, which matters if your target audience is primarily on Google Workspace. For most B2B campaigns, Outlook is the default recommendation.
If this setup exceeds your team's technical capacity, managed infrastructure is a viable starting point. Platforms like Instantly and Smartlead bundle mailbox provisioning, DNS setup, warmup, and sending into one product — collapsing Steps 3 and 4 into a single vendor workflow. Dedicated vendors like Inframail provision Microsoft 365 inboxes at scale if you want the mailboxes managed separately from your sending tool. The tradeoff: faster setup, but your sending reputation sits partly on shared infrastructure rather than domains you fully control.
Step 4: Warm up new mailboxes for 3–4 weeks
New domains and mailboxes start with no sending reputation. Inbox providers treat them with suspicion. Sending campaign volume immediately results in spam placement within days.
Warmup means gradually increasing send volume over time while maintaining positive engagement signals — opens, replies, emails marked as not spam. Tools like Instantly, Lemwarm, or Mailreach automate this by sending mailbox-to-mailbox emails that mimic real human behaviour.
The minimum warmup period before carrying real campaign load is three weeks. Four is safer. Do not rush this. The cost of restarting from a blacklisted domain is higher than the cost of waiting.
During warmup, monitor two things via Google Postmaster Tools: domain reputation (aim for High) and spam rate (keep below 0.1%). If either moves in the wrong direction, reduce volume immediately and investigate before continuing.
Step 5: Build a clean, verified list
List quality is the most underestimated variable in cold email performance. A 10% bounce rate on a campaign will damage your sending reputation faster than any other single factor.
Sources for building your list: LinkedIn Sales Navigator (filtered by your ICP criteria), Apollo.io, Clay (for enrichment and filtering at scale), and intent data providers like Bombora if you're targeting based on buying signals. For most campaigns, Sales Navigator plus Apollo covers the basics.
Before any list enters a sequence, run every email address through a verification tool — ZeroBounce, NeverBounce, or Millionverifier. Remove invalid, catch-all, and role-based addresses. Your target is a verified list with a projected bounce rate below 3%. Above that, the list isn't ready.
Segment by ICP sub-type before you build sequences. A CFO at a 50-person SaaS company gets a different email than a CFO at a 500-person manufacturer, even if the offer is the same.
Step 6: Write sequences built around one insight, not one pitch
The biggest structural mistake in cold email copy is leading with the offer. Recipients don't care about your product in a first email — they care about whether you understand their situation.
A sequence that works has three to four touches over two weeks:
- Email 1 — one specific observation about their situation, one relevant result you've produced for someone like them, one low-friction ask
- Email 2 (day 3–4) — a different angle, shorter, references a concrete case study or number
- Email 3 (day 7–8) — social proof or a different format (a question, a resource)
- Email 4 (day 12–14) — a short breakup email that gives them an easy out and often generates the highest reply rate of the sequence
Subject lines should be 4–6 words, specific to them, and avoid anything that reads like marketing. Personalisation that matters is company-specific or role-specific — not first name insertion.
Step 7: Define the metrics before you launch
The metrics that tell you whether a campaign is working, in order of importance:
- Reply rate — positive and negative combined. Below 2% on a cold campaign means something is wrong: ICP, copy, or deliverability
- Positive reply rate — replies that express interest. Above 1% is a working campaign. Above 3% is a strong one
- Open rate — a proxy signal, not a primary metric. Open tracking can inflate numbers via bot opens. Use it directionally, not absolutely
- Bounce rate — above 3% means your list needs to go back through verification before the next send
- Spam complaint rate — any complaints are a signal worth investigating. More than a handful means something in the copy or targeting is off
Do not scale volume until you have a campaign with a positive reply rate above 1% on at least 300–500 sends. Scaling a broken campaign scales the problem.
The role of other channels
Cold email rarely works in complete isolation for enterprise targets. When a prospect receives your email, the first thing many of them do is look you up — LinkedIn, your website, sometimes a quick Google search. What they find either reinforces your message or undermines it.
This isn't an argument for running LinkedIn outreach in parallel from day one. It's an argument for making sure the supporting surfaces are credible before you send. A thin LinkedIn presence or a website that contradicts your positioning will cost you replies that the email earned.
For campaigns targeting mid-market and enterprise, a LinkedIn touchpoint (a connection request or a comment on their content) in the same window as the email sequence consistently improves conversion. Cold calling on non-responders after two email touches is worth testing if the deal size justifies it.
What a working campaign looks like at 30 days
By the end of the first month, you should have: domains that are warmed and stable, a verified list of 500–2,000 contacts segmented by ICP sub-type, one to two active sequences, and enough sends (300+) to read the data.
If reply rates are below 1%, the diagnosis is usually one of three things: the ICP is too broad, the first email is leading with pitch instead of insight, or deliverability is worse than it looks. Check inbox placement manually before assuming the copy is the problem.
If reply rates are above 2%, the job shifts to scaling the infrastructure to match demand — more domains, more mailboxes, tighter rotation — while keeping the sequences that are working intact.
Cold email is not a broadcast channel. It's a targeting and sequencing problem with a deliverability tax on every mistake you make upstream.
If you're building a cold email campaign and want the infrastructure and targeting right before you send the first email, that's where we start.
