Cold Email Infrastructure Setup: The 2026 Guide to Domains, Inboxes, DNS and Pace

1. Domains: Separate, Forwarded, Boring

Never send cold email from the domain your company runs on. Cold outreach collects complaints, bounces and unsubscribes that no invoice, support thread or sales reply should share a reputation with, and when a sending domain burns you want to retire it, not rebuild your company's mail. Register variants of your brand on .com where you can (yourbrand-mail.com, tryyourbrand.com, getyourbrand.com), forward each one to your real site so a prospect who types it in lands somewhere, and keep the registrations in one account with auto-renew on. Our top-level-domain study is the reason we say .com: on our own fleet, the cheap extensions were refused by Google at rates the .com domains never reached.

One more thing to decide at this step is whether to use subdomains of the main brand instead. They protect less than people hope; the subdomain article covers exactly how much, and the short answer is separate domains for cold, subdomains for transactional.

2. Inboxes: Two or Three per Domain

The ratio we run fleet-wide is two to three inboxes per domain. One per domain spends a domain registration and a warmup on 15 sends a day; five or more means one bad batch takes a large slice of your volume with it. Name them like people (first name, first.last), not like campaigns, and give each a real display name and signature. At three inboxes and 15 a day, a domain carries about 45 cold sends daily; the 300-a-day worked example walks the arithmetic from there.

3. Platform: Match the List

Google Workspace reaches Gmail-hosted prospects with the least friction, Microsoft 365 reaches Outlook-hosted ones, and Azure tenants give Microsoft reach with tenant-level isolation at a much lower per-inbox cost and a much lower per-inbox pace. Run your list through the ESP segmenter before you buy anything: it tells you what share of your prospects sits behind each provider, and that share is what your fleet should mirror. Our matching study has the numbers from our own sending.

4. DNS: Three Records, Set Before the First Send

• •SPF: one TXT record at the root with a single include for the platform, under ten lookups.
• •DKIM: the platform's key published under its selector, and on Microsoft 365, DKIM switched on for the domain in Exchange admin, which is the step most people miss.
• •DMARC: p=none with aggregate reporting to start, p=quarantine once two weeks of reports show only your own mail aligning.

Then send a test message to a Gmail and an Outlook account you control and read the headers. DNS being present is not the check; a delivered message with spf=pass, dkim=pass and dmarc=pass is. The authentication guide has the exact records and the four failures that fill most audits, and the authentication checker runs the whole set in one pass.

5. Warmup: Three Weeks, or Already Done

A new inbox on a new domain has no history, and receivers treat no history as a reason for suspicion. The fix is 14 to 21 days of low, rising, engaged sending before cold volume: a handful a day in week one, climbing to the working pace by week three, with real replies in the mix. Warmup tools automate the correspondence; they do not shorten the calendar. The alternative is prewarmed infrastructure, where the domain and inboxes have already done those weeks and send the day they land. The comparison is honest about what each does and does not do.

6. Pace: The Number That Decides Everything

About 15 cold sends per inbox per day on Google Workspace, about 8 on Microsoft 365, about 2 on Azure inboxes that make it up in count. Spread through business hours in the recipient's time zone, not fired at nine o'clock. Those are not provider limits, which are far higher; they are the paces at which engagement and complaint signals stay below what filters act on, month after month. The previous edition of this guide said one inbox per 50 to 100 daily emails. Fleets run at that pace, and they are rebuilt every quarter.

7. The Pieces Around the Edges

• •Tracking. A shared tracking domain ties your reputation to every other sender on it. Use your own, or turn open tracking off entirely; we did, fleet-wide. The tracking domain checker shows what yours resolves to.
• •Suppression. One global list of opt-outs, bounces and hard noes shared across every domain, inbox and tool, checked before every send. The suppression checker runs a list against it.
• •Bounce rules. Pause a domain at 2% bounces, stop and investigate at 3%. The bounce analyzer turns the codes into a cause.
• •Monitoring. Google Postmaster Tools on every domain with Gmail volume, a weekly blacklist check, and domain expiry watched, because an expired sending domain is a day of bounces you did not plan.
• •A reserve. Keep prewarmed capacity on standby equal to a week of sending. The day a domain burns is not the day to start a 21-day warmup.

What It Costs, and the Shortcut

At the paces above, 1,000 cold emails a day is roughly 47 Google inboxes on 16 prewarmed domains plus 150 Azure inboxes across two tenants, about $514 a month on our pricing. The infrastructure recipes list that plan and five others, every figure checked against the planner at build time, and the calculator prices any volume you type in.

Everything in steps one to five is what we deliver: Google, Microsoft 365 and Azure inboxes on free prewarmed .com domains, forwarded, with SPF, DKIM and DMARC set and verified against a real message before you log in, within 48 to 72 hours of payment. The how it works page is the sequence in full; the inboxing playbook is what we do with the fleet once it exists.