Cold Email Infrastructure: How the Pieces Fit Together

What is cold email infrastructure?

Cold email infrastructure is everything that sits between a list of prospects and a reply landing in your CRM: the sending domains and their DNS records, the mailboxes that send, the tool that schedules and paces sequences, the prospect data, the suppression list, and the handoff into your sales process. It’s tempting to think of it as “the domains and the sending tool.” That’s the visible part. The parts that decide whether outbound is safe and useful are usually the ones missing from the diagram.

This guide walks through each layer, what it’s responsible for, and what tends to go wrong. It assumes B2B outreach to people with a plausible professional reason to hear from you. If the plan is to email anyone whose address you can find, no amount of infrastructure will fix that.

Start with the rules you’re sending under

Before anything technical, decide which rules apply to the people you’re contacting. We’re not lawyers and this isn’t legal advice, but the basics shape the infrastructure, so they come first.

• United States (CAN-SPAM). There is no B2B exemption. Commercial email needs accurate header and sender information, a subject line that reflects the content, a valid physical postal address, a clear way to opt out, and opt-outs honored within 10 business days. The opt-out mechanism has to keep working for at least 30 days after you send. The FTC is also clear that hiring an agency or tool doesn’t move the responsibility away from you.
• United Kingdom (PECR and UK GDPR). PECR’s consent rule for marketing email doesn’t apply to corporate subscribers such as limited companies and LLPs, but sole traders and some partnerships are treated as individuals and generally need consent. Named work addresses are still personal data under UK GDPR, so you need a lawful basis (usually legitimate interests, with a documented assessment), you must tell people where you got their details, and an objection to direct marketing means you stop.
• European Union. GDPR applies, and national e-privacy laws vary a lot. Germany, for example, generally requires prior consent for promotional email to businesses as well as consumers. A single “EU” list is rarely a single set of rules.
• Canada (CASL). Commercial email generally needs express or implied consent, including to businesses. Many cold campaigns should simply leave Canadian contacts out unless there’s a clear basis.

What this means for infrastructure: every sending path needs honest sender identity, a working opt-out, a postal address in the footer where required, and a suppression list that every campaign checks before it sends. Those aren’t features to add later.

The layers of a cold email setup

Sending domains

Cold email is normally sent from domains kept separate from the main company domain. The reasoning is reputation isolation: if an outbound campaign attracts complaints, you don’t want invoices, support replies and newsletters from your primary domain to suffer for it.

Two things are worth knowing. First, separate domains limit the blast radius; they don’t make bad sending acceptable, and mailbox providers can connect related domains. Second, Google counts subdomains toward the primary domain for its bulk sender threshold, so sending from a subdomain of your main domain doesn’t isolate it the way a separate registered domain does. We cover naming, DNS and the setup checklist in the outbound domain setup guide.

Mailboxes

Cold email is usually sent from real mailboxes on Google Workspace or Microsoft 365, not from a bulk email service provider. Each mailbox should belong to a real, named sender who can receive and answer replies. Every mailbox needs a defined daily sending limit, and that limit should start low and rise only when bounce and reply data supports it.

A common mistake is treating mailbox count as capacity. A team adds twenty mailboxes across eight domains and assumes it can send twenty times more. What actually happens is that the same weak list now reaches more people faster, and the reputation problems arrive sooner.

The sending tool

Tools such as Smartlead, Instantly or the sequencing in Apollo connect to those mailboxes, schedule steps, rotate sending across inboxes and stop a sequence when someone replies. They also collect replies into a shared inbox where they can be categorized.

The sending tool is not your system of record. It doesn’t know a prospect became a customer last week unless something tells it. Treat it as an execution layer that reads exclusions from elsewhere and writes outcomes back.

Prospect data

Data quality is the biggest single driver of cold email outcomes we see. Out-of-date addresses bounce. Catch-all domains accept everything and then silently drop it. Role addresses such as info@ go to shared inboxes where a complaint is one click away. Before import, addresses should be verified, the source of each record should be recorded (you’ll need it to answer “where did you get my details?”), and the list should be checked against your exclusions.

Suppression

This is the layer that most often exists only in someone’s head. A proper suppression list is a single place that holds every address and, where appropriate, every domain that must not be contacted: opt-outs, objections, hard bounces, complaints, current customers, open opportunities, competitors and partners.

Every campaign checks it before sending. Every tool that can send email reads from it. And it’s never “cleaned up” by deleting people who asked to be left alone; you keep the minimum needed to keep suppressing them. The ICO’s guidance makes this point directly: keep objectors on a do-not-contact list so you can screen future lists against it.

The CRM

Interested replies need to become CRM records with the right owner, the right lifecycle state and a source that says they came from outbound. The CRM also feeds the sending tool: customer status, open deals and do-not-contact flags should flow outward, so that prospecting never targets an account sales is already working. The outbound CRM handoff guide covers this in detail.

How the pieces connect

Seen as a system, cold email follows the same pattern as any CRM automation: data, state, event, rule, action, measurement.

The data is the prospect list and its source. The state is whether that person is eligible to contact (not suppressed, not a customer, not in another sequence). The event is a send, a bounce, a reply or an opt-out. The rules decide what each event means, for example “a reply saying stop is an opt-out, not a ‘not interested’.” The actions are suppressing, creating a CRM record or routing to an owner. The measurement is bounce rate, opt-out rate, positive replies and pipeline created.

When a team has trouble with outbound, it’s rarely because one tool is wrong. It’s usually because one of those links is missing. Here’s a realistic example.

A sales team buys a fresh export of 4,000 contacts on Monday. The SDR imports it into a new campaign that afternoon. No one checks it against the suppression list in the old sending tool, which the team stopped using in the spring. By Wednesday, a dozen people who asked to be removed months ago have received a new sequence. Two mark it as spam, one replies angrily to the CEO, and the domain’s bounce rate is climbing because the export included addresses that were already known to be dead.

None of that is a deliverability trick problem. It’s an infrastructure problem: suppression lived in one tool instead of in one place.

What to document

Outbound infrastructure changes hands more often than most systems. SDRs leave, agencies rotate, tools are swapped. A short written record prevents most of the damage:

• Each sending domain, its registrar, DNS host, purpose and date registered.
• Each mailbox, its named sender, domain and current daily limit.
• Where the suppression list lives, what feeds it and which tools read from it.
• Which data sources are approved, and how each record’s source is captured.
• How replies are categorized, and which categories create CRM records.
• Who reviews bounce, complaint and opt-out numbers, and how often.

If that list would take you more than an hour to fill in, that’s useful information in itself.

Signs your infrastructure needs attention

Some patterns show up again and again:

• No one on the team can say, without looking, how many sending domains exist or who controls them.
• Opt-outs are handled by marking a reply “not interested” in the sending tool.
• The CRM has outbound contacts with no recorded source.
• Account executives find out about booked meetings from the prospect.
• Volume is increased by adding mailboxes rather than by improving the list.
• Bounce rates are checked only when someone notices replies have dried up.

Each of these is fixable, and most are fixed by process and data flow rather than by a new tool. If you’re seeing several, our guide to cold email deliverability explains how they show up in inbox placement.