Almost everyone sizes a cold email setup the wrong way round. They pick a sending volume, usually a round number that sounds ambitious, and then buy mailboxes to support it. The result is either a setup that costs more than the pipeline it produces, or one that cannot deliver the volume it was bought for.
Work backwards from meetings instead. The chain is short and every link is a number you either know or can estimate honestly.
The chain
Meetings needed → positive reply rate → total replies → reply rate → emails needed → sends per day → mailboxes → domains.
A worked example with conservative numbers. You want 10 meetings a month. If about a third of positive replies become meetings, that is 30 positive replies. If a third of all replies are positive, that is 90 replies. At a 4% reply rate that is 2,250 emails a month, or roughly 110 a working day.
At a safe 25 to 30 sends per mailbox per day, 110 a day is four mailboxes. Not forty. The gap between that and what most teams buy is the entire point of doing the calculation.
The number that governs everything
Sends per mailbox per day. Treat 25 to 30 as the working ceiling for cold traffic on a warmed mailbox. Providers will accept more; reputation will not tolerate it for long, and the failure shows up as placement rather than as an error message.
This single figure is why volume scales through mailbox count and not through pushing individual mailboxes harder. Ten mailboxes at 30 is safe. Three mailboxes at 100 is a domain reputation problem with a two-week fuse.
Mailboxes per domain
Two or three mailboxes per domain, and not more. The reason is containment: reputation is largely a domain-level property, so a domain carrying ten mailboxes concentrates ten mailboxes’ worth of risk in one place. If it goes, it all goes at once.
| Target sends/day | Mailboxes | Domains |
|---|---|---|
| 100 | 4 | 2 |
| 250 | 9 | 4 |
| 500 | 18 | 7 |
| 1,000 | 35 | 13 |
Never send from your primary domain. Use lookalike sending domains and keep the main one for the business, so a bad campaign cannot reach your invoices and your recruiting.
Three adjustments people forget
- Warmup time. New mailboxes need two to four weeks before they carry full cold volume, so a setup bought today produces pipeline next month. Ramp schedules here.
- List size, not just send volume. 2,250 emails a month across a three-step sequence is roughly 750 new prospects a month. If your total addressable market is 4,000 people, you will exhaust it in five months and the plan needs a different shape.
- Reply capacity. Ninety replies a month is manageable. Nine hundred is a full-time job, and unanswered interest is the most expensive waste in outbound.
If your reply rate is under 2%
Do not buy more mailboxes. At a 2% reply rate the constraint is the list or the message, and doubling the infrastructure doubles a bad result while doubling the cost. Fix targeting first, then size the setup against the improved number, it is a much smaller setup than the one you were about to buy.
Next: the longer guide to mailbox sizing, what it costs across the three routes, or how fast you can safely ramp.
Frequently asked questions
How many mailboxes do I need for cold email?
Work backwards from meetings. Ten meetings a month at typical conversion rates needs about 110 emails a working day, which is four mailboxes at a safe 25 to 30 sends each, far fewer than most teams buy.
How many emails can one mailbox send per day?
Treat 25 to 30 cold sends per warmed mailbox per day as the ceiling. Providers accept more, but reputation does not tolerate it, and the damage shows up as placement rather than an error.
How many mailboxes should I put on one domain?
Two or three. Reputation is largely domain-level, so concentrating ten mailboxes on one domain means losing all ten at once if it degrades.
Should I send from my main company domain?
No. Use separate lookalike sending domains so a bad campaign cannot affect the domain your invoices and recruiting depend on.
What if my reply rate is very low?
Do not add mailboxes. Under about 2%, the constraint is the list or the message, and more infrastructure just doubles a bad result at double the cost.