How many mailboxes and domains do you need for cold email?

Start from the audience, not the inbox count

The number of mailboxes is an output. The inputs are how many people genuinely belong in the audience, how many messages each of them will receive, how many days a month you send, and how much volume each sending identity can carry without straining its reputation. Change any input and the right number of mailboxes changes with it.

Mailbox providers publish requirements for authentication, complaint rates, and unsubscribe handling, but not a per-mailbox quota for cold outreach. Google’s guidance is to start with low volume, send at a consistent rate, avoid bursts, and increase slowly. Any per-mailbox number you plan with is therefore your own risk decision, and it should be written down as one. [1]

That is why a rule such as “three inboxes per domain, fifty emails each” travels badly between campaigns: it hides the assumptions that make it true or false. Our sending-domain guide makes the same point. Capacity comes from the audience, a safe schedule, and operational reserve—never from a formula copied from someone else’s campaign.

The capacity model, step by step

Actually Agentic’s quote follows the steps below, using the operator’s configured values. The example assumes a 5,000-contact ICP, which is an estimate from the customer’s agent rather than a verified list. [2]

Messages per lead is the input people forget. A four-step sequence does not cost four messages per lead, because replies, permanent bounces, and unsubscribes stop the later steps. The current default plans for 2.2. If every lead received all four steps, the same mailbox would reach only 100 new people a month.

Coverage months decides how quickly the audience is reached. The default recommendation covers the estimated ICP in about six months. A shorter period needs more mailboxes and domains for the same list; a longer one needs fewer and leaves time to learn from early replies before most of the audience has heard from you. On the scale screen, the full-ICP recommendation is the starting point and the customer can dial the mailbox count down, which lengthens the coverage period without raising any mailbox’s limit.

Which assumptions move the answer most

Run the numbers with cautious inputs first. If the cautious version produces a domain count that looks absurd, the audience is probably too broad—a targeting problem, not an infrastructure one.

• Messages per mailbox per day: cutting the default from 20 to 10 drops each mailbox from 181 to 90 new leads a month and roughly doubles the pool.
• Sending days: sending on all 30 days of a month instead of about 20 business days raises one mailbox to 272 new leads, at the cost of weekend sends outside typical business hours.
• Messages per lead: a longer sequence with fewer early replies pushes this up; at 3 messages per lead, one mailbox reaches 133 new leads instead of 181.
• Coverage months: reaching the same audience in three months instead of six roughly doubles the recommended mailboxes.
• Mailboxes per domain: this changes the domain count, not lead capacity. Fewer per domain spreads identity risk across more domains but adds registration, DNS, and renewal work.

Why the first month is smaller than the model

The formula describes a mailbox at its planned steady state, and new identities do not start there. Actually Agentic warms each new managed domain for at least the configured minimum—14 days by default—before campaign traffic, and domain, mailbox, DNS, and blocklist checks must pass again before a person approves launch. [2]

Campaign volume then ramps. Each mailbox starts at five campaign messages a day and adds three per day until it reaches its approved limit, so a 20-message mailbox reaches full volume about five days after its ramp begins. No plan can take a mailbox past the configured ceiling, 25 messages a day by default.

So a first calendar month usually delivers well under the steady-state figure. Set expectations around the launch date. Adding mailboxes to compensate multiplies the number of new, unproven identities rather than making any one of them more trusted.

Budget, caps, and supply also bound the result

Sending capacity is one ceiling among several. The quote also limits monthly lead checks to what the approved budget covers after fixed costs and an operating reserve, and one campaign is capped at 50,000 lead checks a month. When the budget is the binding limit, the quote says how many of the requested leads it supports instead of quietly shrinking the plan. [2]

Supply is the ceiling most calculators ignore. Monthly capacity is how many candidates the system can discover and verify, not a promise that the audience contains that many reachable people. Native discovery is bounded by the company seeds the agent supplied and by what those companies publish; a larger program needs a reviewed CSV import or a narrower plan. [2][3]

The agent must also propose enough plainly related domain names for the domains the plan needs. If it did not, the plan goes back for more choices rather than the system inventing names on its own.

Leave headroom for pauses

A pool sized exactly to the audience has no slack. Every Actually Agentic mailbox keeps its own daily ceiling, so when capacity drops—a mailbox is unhealthy or a campaign pauses for review—throughput falls. The remaining mailboxes are never pushed past their limits to make up the difference.

If the plan only works when every mailbox runs at its maximum on every sending day, shrink the audience or extend the coverage period instead. A campaign that needs perfect conditions to keep its schedule will eventually miss it.