Cold email infrastructure: how to build it in house

Building a cold email system in house sounds simple: buy a domain, make a few inboxes, load a tool. Then reality turns up. DNS records that fail quietly, a warmup period nobody planned for, a list that bounces, replies scattered across ten mailboxes. This post lays out the full cold email infrastructure stack, what each piece needs, what it costs in time, and when handing it to an agency is the better call.

Cold email infrastructure, piece by piece

1. Sending domains, separate from your main one

Cold email goes on its own domains, never the one your team uses for customers, invoices and press. Outbound carries more risk of bounces and complaints than normal mail, and reputation attaches to the domain. If a campaign burns a sending domain, you retire it and your real mail is untouched.

Typical practice is a lookalike of your brand (a different top-level domain or a small prefix or suffix), pointed at your website so a curious prospect finds a real company. Keep a list of which domain belongs to which campaign, because you'll need it when something goes wrong.

2. Inboxes per domain

Each domain needs several inboxes, and each inbox sends a limited number of emails a day, so volume is spread across many senders instead of piled on one. How many inboxes per domain, and how many emails per inbox, are decisions you make from your volume target. How many cold emails per day shows how to work that out.

Use Google Workspace or Microsoft 365 for the mailboxes. Both document their own sending limits, but those are ceilings, not targets.

3. DNS: SPF, DKIM, DMARC, and the rest

Every sending domain needs SPF, DKIM and a DMARC record before it sends anything. Google's sender guidelines and Yahoo's sender hub both require this. You also need working MX records so replies reach the inbox, and, if you track opens or clicks, a tracking domain you control. This is the step that quietly fails most often, and it multiplies: ten domains means ten sets of records. SPF, DKIM and DMARC for cold email covers each one.

4. Warmup

A new domain and a new inbox have no sending history, and providers are wary of anything that goes from nothing to full volume. Warmup builds the history first, and then you ramp slowly. Plan for that wait before a domain carries a full campaign; it is not a one-day job. Email warmup explains the process.

5. A sending tool

You need something that sequences emails, rotates senders, respects daily caps, stops a sequence on reply and shows you what happened. In SendingOS, email, LinkedIn and calling run in one sequencer, and every inbox's connection is checked every few minutes; an inbox that fails stops sending until it passes again. Whatever you pick, check how it handles a failing inbox, because that is where in-house systems leak.

6. List source and verification

The list decides more than the copy does. Decide where the leads come from and how fresh they are, then verify every address before it enters a campaign. Bad addresses bounce, and bounces damage the domain you just spent weeks warming. In SendingOS, email verification is free, and waterfall enrichment runs through Airscale at $0.005 per verified email and $0.10 per verified phone number, charged only for verified results.

7. Reply handling

Replies are the whole point, and they arrive in every inbox you own. Someone has to read them, label them, answer the interested ones fast and stop the sequence for everyone who replied. In SendingOS, all inboxes and channels feed one Unibox, replies are labelled for you (Interested, Meeting, Out of office) and a reply on any channel stops the sequence everywhere. Without a shared inbox, this becomes someone's full-time job of logging into mailboxes.

8. Monitoring

Domains drift: bounce rates creep up, complaints rise, one provider starts filtering you. Someone needs to look at the numbers every day and act before the damage spreads. That means bounce reasons, Postmaster Tools for Gmail, and placement checks. Email spam test describes the checks.

What it costs in time

We won't invent dollar figures, because they depend on your provider choices. The time cost is easier to reason about, and it comes in three parts.

• Setup. Buying domains, creating inboxes, publishing records for each domain, connecting everything to the tool. Done by hand it is repetitive, and mistakes compound across domains.

• The warmup wait. Not hands-on time, but a delay of weeks before you can send at volume. It also means you must plan capacity ahead of when you need it.

• Ongoing care. Daily monitoring, replacing burned domains, refilling lists, reading replies. This never ends, and it is the part people underestimate.

Add up the weekly hours honestly, and ask who owns them. A system nobody owns degrades.

When an agency makes more sense

Building in house makes sense when outbound is core to how you grow, you have someone who will own it, and you can wait out the setup. An agency (or a managed service) makes more sense when:

• you need meetings from month one and can't wait through a build and warmup;

• nobody on the team has time to watch deliverability daily;

• you're testing whether cold email works for you at all, and don't want to commit to infrastructure first;

• you run outbound for several clients and need each one's sending kept apart. In SendingOS, every client workspace sends on its own isolated IPs, so one client's bad list lands on that client's IPs, not another's.

If you do build, start with one domain's worth of the stack, prove it end to end, then copy it. The whole deliverability picture shows what to measure at each stage.