A sales manager drops a screenshot of the daily send cap into Slack, with the division already done underneath. One mailbox is clearing about 40 first touches. The number on the forecast is closer to 150. They want another mailbox: a new Google Workspace user, a Microsoft 365 mailbox on a domain registered the same week, or a vendor bundle described as aged and warmed and ready for the sequencer before standup ends.
Sequencers are built to spread volume across mailboxes, so the screenshot feels like a plan. A new seat is a small line item next to an argument about the offer or the list. Leadership asked for more first touches, and a second mailbox changes that count before the next forecast call. The extra count helps only when mailbox one has already shown that strangers reply to this mail on purpose.
Leave the daily cap alone
For cold outbound, one mailbox that settles around a few dozen new touches a day, once it is in real use, is a sane place to operate. Providers differ, older domains tolerate more than new ones, and a mailbox that gets replies can carry more than a mailbox that gets ignored. The order stays the same either way. Find out whether the program works on one mailbox. Add another only after one week of sends looks like the week before it.
Teams usually flip the order. The warm-up chart turns green because a network of other mailboxes has been trading empty notes with yours. That chart measures mail the warm-up network exchanged with itself. I have seen a mailbox finish that exchange and then hit spam the week a real list was attached. There is nothing to copy until mailbox one has sent live first touches, you have checked where they landed, and you have read the replies.
Give it a few weeks of that live mail before anyone orders a second seat. Count the messages a stranger received, carrying the offer you intend to keep running, on a list you would defend in a pipeline meeting. Warm-up exchanges and notes to coworkers sit outside that sample.
Keep the hard bounce count for those sends somewhere the team can open without a scavenger hunt. Put the spam complaint count beside it, including the report inside the ESP that nobody has opened since the account was created. Separate human replies from automatic ones. Out-of-office notes and mailbox-full warnings inflate a reply rate when every response is scored as interest. In the same week, deliver the live template to a Gmail account you control and to a Microsoft 365 account you control, then write down whether it lands in the inbox, in promotions, or in spam. A placement slide from onboarding does not replace opening those accounts.
If those figures are blank, another mailbox spreads the same unknown across more sends. Hold the current cap until someone can fill the counts in from this week's mail.
Same domain or a new one
Adding a user on the domain you already send from looks like the careful choice. sam@ gets a sibling, or a new hire's address is pulled into cold mail because the license is already on the card. People treat the domain's age as if it transfers from one mailbox to the next. Receivers score the domain, and they notice a fresh stream of unsolicited mail on it. Complaints on the second mailbox accumulate against the reputation the company already uses for customers, invoices, and the founder's own correspondence. Cold mail from a secondary mailbox can push that other mail into spam. The outbound test and the customer mail were on one domain the whole time, whatever the admin console showed.
A separate outbound domain is a reasonable design once mailbox one is producing conversations you would want more of. It keeps cold sending off the domain customers already trust. Used in week two, it skips the evidence those weeks were supposed to collect. A domain registered this month has a thin history with Gmail and Microsoft, and those filters treat new sending domains cautiously. Point it at the same file and the same offer mailbox one could not turn into replies, and you now have two domains with the same complaint problem. Buying the name on Friday and loading sequences on Monday puts live mail on a domain that has not been warmed.
An alias stays on that same domain. outreach2@ is another From line on a sender the filters already know. If they have started to doubt the domain, the local part does not give you a different reputation.
Rented mailboxes need a colder reading. A vendor selling aged domains with mailboxes already warmed should be able to say where those domains have sent, who else is sending on them this month, and which inbox receives the message when a prospect hits reply. Vague answers mean your offer is going out on a reputation you cannot inspect. When that vendor has a bad week, or when the contract ends, the replies and the complaints are still attached to infrastructure you do not administer. Sending from a domain you own, through an ESP you can log into, at a volume you can explain to a customer who missed an invoice, takes longer this month. Next month you can still operate it without asking for access.
A second mailbox will not clean the file. Cutting a stale export in half and assigning each half to its own mailbox does not verify addresses, drop role accounts, or exclude people who already told you to stop. Do that work on the file you already have. Mailbox two should not inherit the rows mailbox one already bounced.
Ramp one mailbox, then stop
I add the next mailbox after the first one has been uneventful. A few weeks of real first touches are in the books. Someone opened the seed accounts this week and mostly found the inbox. A person on the team can state the recent hard bounce count and the complaint count from a note written this week, without a support ticket to locate them. Complaints have been rare enough that a new one would stand out. The replies include people who understood what was asked, and some of those people declined. That pattern is healthier than a run of "who is this" and removal demands. Follow-up is keeping pace with those replies. If mailbox one already produces more live conversations than the team works the same day, another mailbox extends the backlog.
Settle DNS on the sending domain before the new seat is purchased. SPF has to authorize the system that hands the message to the receiving server. DKIM has to sign. DMARC should be publishing, on a monitor policy if you are not ready to reject, so alignment problems arrive as reports. The domain in the From header should be the domain that signs. Adding the mailbox leaves the signing mismatch in place. The display name can still imply the corporate domain while the From address sits on a lookalike.
Start the new mailbox low, around 10 sends on a normal day, and hold that level for several days. Move toward 20 only if bounces, complaints, and the seed checks still resemble mailbox one. Read the domain totals. A per-mailbox chart can make both mailboxes look calm while the domain is doing 90 of the same email. A day of 30 careful emails can look ordinary to a filter. A day of 90 near-copies can look like a burst, even when each mailbox stayed under its own cap.
The reason teams skip the wait is usually a public meeting target and a short week. The second mailbox is a way to change the send chart before the forecast call. I would rather walk into that call light on volume than spend the following month explaining why mail from the main domain failed checks it was passing in the spring.
Before the seat goes on the card, write down the bounce count, the complaint count, and where yesterday's template landed in the seed accounts. Bring that note to the meeting where someone asks for the mailbox.