Scaling cold email infrastructure: the sender pool
How a sender pool of 100 inboxes across 34 domains scales cold email outbound: rotation, redundancy, warmup, and why 25 inboxes in an error state is normal.
The truth about scaling cold email: redundancy beats perfection
Most cold email guides stop exactly where it gets interesting: at the first cleanly configured domain. SPF, DKIM, DMARC set, a few weeks of warmup, first campaign. That works as long as you send a few hundred emails a month. The moment you want to scale seriously, the logic flips. It is no longer the single perfect inbox that matters, but how your entire sender infrastructure copes with failure.
Pile more emails onto each inbox as volume rises and you burn domains faster than you can buy new ones. The mistake is treating scaling as a volume problem instead of an architecture problem. This article shows what a cold email infrastructure built for volume looks like, drawing on real numbers from a live workspace rather than best-practice folklore.
The technical groundwork comes first (deliverability basics and a proper warmup). This article is about the layer above: the pool.
The sender pool: the actual building block
Scalable infrastructure is not one domain with many mailboxes, it is a pool of many domains with a few inboxes each. The figures from our own setup, as of June 2026:
| Metric | Value |
|---|---|
| Total sender inboxes | 100 |
| Lookalike domains | 34 |
| Inboxes per domain | 3 to 5 |
| Provider mix | 44 Google, 40 Microsoft, 16 IMAP |
| Campaign emails per inbox per day | ~15 |
| Campaigns analysed | 47 |
| Emails sent | 87,253 |
| Overall bounce rate | 3.08% |
| Leads contacted | 27,829 |
The principle behind it: scale horizontally, never vertically. Not more volume per inbox, but more inboxes. At around 15 real cold emails per inbox per day, 30 inboxes already produce 450 clean emails daily. Push 30 inboxes to 50 emails each instead and you produce the same number, but at a fraction of the deliverability and a multiple of the domain risk.
The main domain stays out of it. You send exclusively through lookalike domains, for the main domain firma.de variants such as firma-ai.com or firma-labs.com. If one of them dies, at most 5 inboxes are affected, not your business communication.
Redundancy is the foundation, not the insurance
Now the part no glossy guide mentions: in our workspace, 25 of 100 accounts sit in an error state at times. So a quarter of the pool is regularly not ready to send, because of reauthentication, provider limits or temporary blocks.
That sounds like a problem and is in truth proof that the architecture is built right. With 100 inboxes across 34 domains, no single outage breaks the campaign. The remaining 75 inboxes keep sending, volume redistributes, nobody notices. That is exactly the difference between a setup optimised for perfection and one optimised for operation.
Plan your infrastructure so that nothing may fail and you do not have infrastructure, you have a single point of failure with extra steps. Redundancy means the failure is priced in. Hence 3 to 5 inboxes per domain, and many domains rather than a few large ones.
The provider mix as risk spreading
A pool from a single provider is fragile. Microsoft and Google change their delivery algorithms two to three times a year, often without notice. Run your entire outbound through Microsoft and one of those shifts can halve your pipeline overnight.
Our mix of 44 Google Workspace, 40 Microsoft and 16 IMAP inboxes spreads that risk. It also carries a delivery advantage: on the recipient side the same two providers dominate, and mail from Google to Google or Outlook to Outlook is delivered structurally better. The mix is therefore insurance and performance lever at once.
Warmup never stops
When scaling, one rule surprises many people: warmup runs permanently on all accounts, not only the new ones. All 100 inboxes have warmup active. The reason: warmup traffic keeps an inbox’s engagement signals stable, even when a campaign pauses or volume fluctuates.
New inboxes go through around three weeks of warmup before the first campaign send, with an account limit of 20 per day. Established accounts sit at a limit of 100 per day but still send only the conservative ~15 real cold emails, the rest is warmup and buffer.
The silent killer: list hygiene
The best infrastructure in the world does not survive a bad list. The proof comes from two campaigns in the same workspace, with the same domains and the same setup:
| Campaign | Bounce rate |
|---|---|
| Validated list (regional, checked) | 0.4% |
| Unvalidated import | 7.7% |
A factor of 19, from email validation before import alone. This is no footnote: domain damage starts at around 5% bounce, and above 8% the domain is effectively burned. A single unvalidated list can destroy inboxes that took weeks to build. That is why we validate every list before import, targeting a bounce below 2%.
How you measure success when you scale
Open rate is not a metric in this setup, it simply does not exist. Across all 47 campaigns, open and link tracking are disabled, because tracking pixels and rewritten links are among the loudest spam signals and actively cost deliverability at high volume. We measure what counts:
- Unique replies per contacted lead (3.9% on average, 11.9% for the best campaign with a regional, sharp ICP)
- Qualified replies with genuine buying interest
- Bounce rate as an early-warning system, alarm above 3%
Conclusion: infrastructure is an operational decision
Scaling cold email infrastructure means shifting from perfection to operation. It is not the one indestructible inbox that carries the volume, but a pool in which something is always failing and everything still keeps running. 100 inboxes across 34 domains, mixed providers, permanent warmup, hard list hygiene and measuring by replies instead of opens: that is not a trick, it is an architecture that prices in failure.
The catch: this setup is full-time work. Registering and maintaining domains, keeping DNS correct, monitoring warmup, chasing error states, validating lists, balancing volume across the pool. This is exactly the infrastructure GTM Goat, CegTec’s outbound system, runs as a managed layer: across real campaigns it steered on the order of 87,000 emails, at 3.08% bounce across the whole pool. You get qualified meetings, not the job of babysitting a sender pool. What that looks like for a scaling vertical like B2B SaaS is shown on that page.
If you want to see the operation live rather than build it yourself: GTM Goat can be tested free for four weeks, including running infrastructure from day one.
Common questions
How many inboxes and domains do I need for scalable cold email outbound?
As a rule of thumb from a live DACH setup: around 3 to 5 inboxes per domain, spread across many lookalike domains. Our own workspace runs 100 inboxes across 34 domains. You scale horizontally through more inboxes, not vertically through more volume per inbox. At roughly 15 campaign emails per inbox per day, 30 inboxes already mean 450 clean cold emails a day.
Why should I mix several email providers instead of only Google or only Microsoft?
Because provider diversification spreads failure risk. Our pool holds 44 Google Workspace, 40 Microsoft and 16 IMAP inboxes. Microsoft and Google change their delivery algorithms two to three times a year. Send on a single provider and one such shift wipes out your entire outbound. A mixed pool cushions that, and mail from Gsuite to Gsuite or Outlook to Outlook structurally delivers better.
Is it normal for sender inboxes to fail?
Yes. In our workspace, 25 of 100 accounts sit in an error state at times. At this scale that is normal, not an alarm signal, and it is exactly why you build redundancy. Because only 3 to 5 inboxes hang off each domain, a domain outage hits at most 5 inboxes and the campaign keeps running on the others. Redundancy is not a nice-to-have, it is the foundation of the architecture.
How much does list hygiene affect the scaled infrastructure?
Massively. In the same setup, with the same domains, a validated list bounced at 0.4% and an unvalidated one at 7.7%. That is a factor of 19, from email validation before import alone. Domain damage starts at around 5% bounce, and above 8% a domain is effectively burned. A bad list destroys the very infrastructure you built with such effort.
Why do you measure by replies when scaling, not by open rate?
Because open and link tracking are among the loudest spam signals and actively cost deliverability at high volume. Across all 47 analysed campaigns tracking is disabled, so open rate does not exist as a metric. We measure unique replies per contacted lead (3.9% on average, 11.9% for the best campaign), qualified replies, and bounce rate as an early-warning system above 3%.