Outbound infrastructure architecture for a B2B team
What domains, inboxes, warmup, authentication and monitoring you need to run outbound safely, how they fit together and how many of each your volume requires.
ON THIS PAGE · 9 SECTIONS
- What outbound infrastructure actually is
- The layers, in the order they get built
- How many domains and inboxes you actually need
- The authentication records that gate delivery
- Warmup and ramp, step by step
- The LinkedIn layer runs on the same logic
- Building it without overspending
- The monitoring that keeps it healthy
- Questions about outbound infrastructure
THE SHORT VERSION
- Outbound infrastructure is the domains, inboxes, authentication, warmup and monitoring under your list and your message, deciding whether either one gets delivered.
- Google and Yahoo have required SPF, DKIM, DMARC and a spam complaint rate under 0.3% on bulk senders since February 2024. Both now enforce it automatically.
- Plan on 2 to 3 inboxes per sending domain and 20 to 30 safe sends per inbox per day once warmed, far less in the first month.
- LinkedIn infrastructure runs on the same logic: a warmed account, a slow ramp and a cap of roughly 15 to 25 connection requests a day.
- Build the sending layer to match your volume. Oversized infrastructure sits idle and ages badly. Undersized infrastructure gets throttled within weeks.
Outbound infrastructure is the sending layer underneath a GTM system: the domains, inboxes, authentication records, warmup schedule, LinkedIn accounts and monitoring that let an email or a connection request reach a real inbox instead of a spam folder or a restriction notice. It is not your account list and it is not your message. It is the plumbing that decides whether either one ever gets seen. A good list with good copy still fails on a sending setup that mailbox providers do not trust.
This piece covers the layers in order, how many domains and inboxes a given volume actually needs, the authentication records every major mailbox provider now checks before delivery, the warmup schedule that keeps a new domain or LinkedIn account out of trouble and the monitoring that catches a problem before it costs you a week of sending.
What outbound infrastructure actually is
Teams that are new to outbound tend to think of infrastructure as a side task: buy a domain, make a few inboxes, hook up a sending tool, done. Treated that way it becomes the reason campaigns stall. A sending domain carries a reputation score with every major mailbox provider, built from how recipients treat your mail over time. That score is invisible until it turns against you, usually as mail routed to spam or an account placed under a sending restriction with no explanation.
The job of outbound infrastructure is to keep that reputation score high enough that your list and your message get a fair chance to be read. It sits between the account list you built and the reply that lands in a rep's inbox. If you have already mapped your market, our piece on how we map a market before the first email goes out covers that step. Infrastructure is what carries that list to a real person once it exists.
The layers, in the order they get built
Outbound infrastructure is not one thing. It is a stack of separate layers, each with its own failure mode. Each one depends on the layer below it being healthy first.
| Layer | What it does | How it fails |
|---|---|---|
| Sending domains | Carries the reputation your mail is judged on, kept separate from your company website | Gets filtered after one bad campaign, taking your real domain down with it if you reused it |
| Inboxes | The mailboxes that actually send and receive replies | Flagged for automation before anyone has written a word of copy |
| Authentication | SPF, DKIM and DMARC prove you are who the From address says you are | Mail routed straight to spam, often with no bounce notice at all |
| Warmup | Builds a sending history before real volume starts | A brand new domain sending full volume on day one, flagged within days |
| Sending tool | Schedules sends, rotates inboxes and tracks replies against per inbox caps | Manual sending that cannot rotate inboxes or hold a daily limit |
| LinkedIn accounts | A second channel with its own reputation and its own limits | Restricted or put into a connection request slowdown for weeks |
| Monitoring | Watches spam complaint rate, bounce rate and inbox placement every week | A problem gets caught only once replies disappear, often two weeks after it started |
Each layer depends on the one above it. Skipping warmup to save two weeks usually costs a domain within a month.
How many domains and inboxes you actually need
Most teams either guess too high and buy ten domains they never fill or guess too low and throttle their own volume within a month. The right number comes from the volume you plan to send, not from a vendor's default package.
Source: Scaled Mail, 2026
Work the math from your target, not from a round number. Say you want to reach 2,000 net new prospects a month on email. At 25 safe sends per warmed inbox per day across roughly 20 working days, one inbox carries about 500 sends a month. That is four inboxes. Capped at two to three inboxes per domain, four inboxes split cleanly across two domains, each carrying two inboxes, with headroom to add a third inbox per domain before you need a fifth domain. Build to that number, then revisit it once you have a few months of real send data instead of a vendor estimate.
The account list drives this number as much as the message does. If your mapped market only supports 400 accounts worth targeting this quarter, two inboxes on one domain can carry that volume on its own. A bigger build just sits idle while it ages. Match the stack to the list, not the other way around.
The authentication records that gate delivery
Since February 1, 2024, Google and Yahoo automatically block mail from any bulk sender, defined as more than 5,000 messages a day to personal accounts, that does not meet a fixed set of authentication rules. This is not a best practice anymore. It is a hard gate that most major mailbox providers now enforce without a human reviewing the decision.
- SPF record published for every sending domain, naming the servers allowed to send on its behalf.
- DKIM signing enabled and verified, with a passing signature on outgoing mail.
- DMARC policy published at a minimum of p=none, with the visible From domain matching the domain authenticated by SPF or DKIM.
- One-click unsubscribe header present on every bulk message, honoured within 48 hours of a request.
- Spam complaint rate tracked weekly, kept under 0.3% and ideally under 0.1%.
Source: Mailgun, 2026
The spam complaint threshold is the one teams underestimate. A rate of 0.3% sounds generous until you realise it means three complaints per thousand sends. A single disengaged list segment can push a whole domain past it inside a week. Below 0.1% should be the actual target, with 0.3% treated as the point of no return rather than a safe ceiling.
Warmup and ramp, step by step
Warmup exists because mailbox providers trust a sending history, not a sending tool. A domain or inbox with no history looks identical to a spammer's burner account, so every new domain and every new inbox needs to earn trust before it carries real volume.
- Register and configure DNS. Buy the sending domain, point MX records at your mailbox provider and publish SPF, DKIM and DMARC before a single mailbox goes live.
- Let the mailbox sit and receive. A few days of normal receiving activity before any sending gives the domain a baseline that is not purely outbound.
- Start warmup at low volume. Begin around 5 to 15 sends a day, either through a warmup tool that trades mail with other warmed inboxes or through genuinely varied, low volume manual sends.
- Ramp over 2 to 4 weeks. Increase volume gradually rather than jumping from 10 to 60 in a day, which is the single clearest pattern mailbox providers flag as automation.
- Connect to real campaigns at steady state. Only after 14 or more days of ramp should an inbox take on real cold volume, at the 20 to 30 a day ceiling, not before.
A domain warmed this way before any real sending starts is the single biggest factor separating teams whose deliverability holds for a year from teams who burn a domain every quarter and wonder why.
The LinkedIn layer runs on the same logic
LinkedIn infrastructure gets treated as an afterthought next to email, which is backwards, since a restricted LinkedIn account is just as dead as a blacklisted domain and takes about as long to recover.
Source: Salesforge, 2026
Two things matter more than the raw numbers. First, a sudden jump in volume reads as automation faster than a high but steady volume does, so a consistent 20 a day beats an inconsistent 10 one day and 40 the next. Second, LinkedIn penalises a low acceptance rate harder than it penalises volume, so a badly targeted list of connection requests gets an account restricted long before it hits any daily cap. The fix for that lives upstream of infrastructure, in the account and persona targeting covered in our piece on B2B messaging by segment.
Building it without overspending
The common failure runs in both directions. Some teams buy a large domain and inbox package on day one because a vendor sells it that way, then let half the inboxes sit unused while reputation quietly ages without ever carrying mail. Others start on a single personal inbox, hit the 20 to 30 a day ceiling within a week and conclude that outbound email does not work for them.
THE USUAL WAY
- Buy a fixed package of domains and inboxes before volume is known.
- Skip warmup to start sending sooner.
- Check deliverability only after replies stop coming in.
THE SYSTEM WAY
- Size domains and inboxes off the account list and target volume.
- Warm every domain and account for 14 or more days before real sends.
- Track spam complaint rate and inbox placement every week, not after a drop.
A sending domain does not fail the day you notice it. It fails two weeks before that. You only find out when the replies stop.
FROM THE YARD
Keep one spare, pre-warmed domain ready at all times once you are sending real volume. The moment a domain's spam complaint rate climbs past 0.1%, move the campaign to the spare while the flagged domain rests, rather than pushing through on a domain that is already losing trust.
The monitoring that keeps it healthy
Infrastructure is not a one time build. A domain that passed every check at launch can still degrade three months later because a list segment went stale or a campaign ran hotter than planned. The teams that keep deliverability steady for a year run the same short checklist every week rather than waiting for a visible problem.
Check the spam complaint rate per domain against the 0.1% target, not the 0.3% enforcement line. Check the bounce rate, since a climbing bounce rate usually means a list has gone stale faster than expected, a problem covered in more depth in our guide to waterfall enrichment and how verified data keeps that rate low in the first place. Run an inbox placement test monthly to see whether mail is actually landing in the primary inbox rather than a promotions tab or spam folder. Check domain and IP blacklists, since a listing can happen from something outside your control, like a shared IP range another sender poisoned.
Honestly, no amount of monitoring fixes a list that should not have been targeted in the first place or a message nobody wants to read. Infrastructure protects delivery. It does not manufacture interest. If replies are low despite clean deliverability metrics, the problem has moved upstream into targeting or the message itself. No amount of domain rotation will fix that. We cover how that upstream layer connects to the sending stack in the system we build with clients.
Questions about outbound infrastructure
How many sending domains does a small team need to start?
A team sending a few hundred emails a month can often run on a single sending domain with two inboxes, as long as both are properly warmed. Scale domains up only once you can see real send volume data, not before.
Should outbound email ever send from the company's main domain?
No. Cold outreach carries deliverability risk that a secondary domain contains. If a sending domain gets flagged, you retire it and warm a new one, while invoices, proposals and support mail on the main domain stay untouched.
How long does it take to warm a new domain?
Plan on at least 14 days of gradual ramp before a domain carries full cold volume, starting around 5 to 15 sends a day and increasing gradually. New domains pushed to full volume on day one are the most common cause of a burned domain in the first month.
What happens if a domain's spam complaint rate goes over 0.3%?
Google and Yahoo automatically filter mail from senders above that threshold. Recovery is slow even after the underlying list or message problem is fixed. Treat 0.1% as the real ceiling and move volume to a spare, pre-warmed domain the moment a domain crosses it.
Does LinkedIn infrastructure need the same care as email?
Yes. A LinkedIn account needs the same slow ramp, consistent daily volume and attention to acceptance rate that a sending domain needs. A restricted account takes as long to recover as a blacklisted domain and a poorly targeted list causes restrictions faster than high volume does.
Sources: Mailgun, Yahoogle bulk sender requirements, 2026, Scaled Mail, inboxes per domain for cold email, 2026, Salesforge, LinkedIn connection request limits, 2026