How many cold emails can you send from one mailbox per day?

TL;DR

• Safe cap: 30-50 cold sends per mailbox per day. Below provider limits, but high enough to actually scale.
• Google Workspace allows 2,000/day, Microsoft 365 allows 10,000/day — but these are bulk limits. Sustained cold sends at those rates will burn the mailbox in days.
• Warmup the mailbox for 21-28 days before any cold sends. Start at 10/day on day 1, ramp to 30-40/day by day 21.
• To hit 1,000 sends/day: you need ~25-30 mailboxes, ideally spread across 10-15 sending domains.
• Spread sends across the workday. 30 sends compressed into one hour reads as a blast. 30 sends over 8 hours reads as a human typing.
• Weekends and overnight sends are bot signals. Send during recipient business hours.

Provider limits vs reputation limits — different things

There are two ceilings on how many cold emails you can send from one mailbox per day. Most operators only think about the first one. The second one binds tighter.

The provider ceiling: Google Workspace, Microsoft 365, and other providers cap raw send volume per mailbox per 24-hour window. Google sits at 2,000 recipients/day. Microsoft 365 is 10,000/day. These are infrastructure caps designed to prevent a compromised account from blasting spam at scale.

The reputation ceiling: Gmail, Outlook, Yahoo, and corporate mail gateways watch sending behavior and assign reputation per sender. A mailbox that sends 500 cold emails in one day looks like a marketing blast (because it is one). Spam classifiers respond accordingly. Inbox placement drops, spam complaint rates rise, and the mailbox is operationally dead within 5-10 days even though the provider hasn’t suspended it.

The reputation ceiling is far lower than the provider ceiling. For cold outbound, where every send is to a recipient who didn’t ask for the message, the safe sustained cap is roughly 30-50 sends per mailbox per day. That’s the number you should plan against. The 2,000/day provider limit is irrelevant to your operating model.

Why 30-50 per day is the practical safe cap

The 30-50 number isn’t arbitrary. It comes from how mailbox providers infer “real person” versus “bulk sender” behavior:

• Engagement-to-volume ratio. A real human sending 30 emails a day expects most of them to be replied to. A bulk sender sending 500 a day expects 1-3% reply rate. Providers infer the difference and weight the mailbox accordingly.
• Spam complaint amplification. At 30 sends/day, a 0.1% complaint rate is roughly 1 complaint per 1,000 sends — invisible noise. At 500 sends/day, the same complaint rate is 1 complaint every other day, and the absolute volume triggers Gmail and Microsoft’s reputation systems.
• Recipient-domain coverage. 30 sends typically land at 28-30 different recipient domains. 500 sends land at 200-400. Receiving providers compare notes; a sender hitting 400 distinct domains in one day with low engagement is exactly the pattern they’re built to throttle.

The empirical pattern across teams running production cold outbound: 30-40 sends per mailbox per day is the sustainable steady state. 50 is the ceiling for a well-warmed mailbox with high engagement. Above 50 sustained, you start eating reputation faster than you build it.

The warmup ramp — why you can’t start at 30

A brand-new mailbox cannot send 30 cold emails on day 1. It has no sending history, no engagement record, and any new mailbox sending 30+ messages in its first week is overwhelmingly likely to be spam (because most are). The ramp is the difference between a usable mailbox and a burned one.

The standard warmup curve, over 21-28 days:

• Days 1-7: 5-15 sends/day, mostly to warmup pool addresses (mailboxes you control) and very-low-risk recipient lists. Replies, opens, and clicks build up the engagement record.
• Days 8-14: 15-25 sends/day, mixing warmup-pool engagement with the first cautious cold sends to vetted, low-spam-risk recipients.
• Days 15-21: 25-35 sends/day, mostly real cold outbound now, with warmup engagement still running in the background.
• Day 22+: 30-50 sends/day sustained.

See inbox warmup for the full engagement-curve and synthetic-warmup-detection discussion. The headline: automated warmup tools that simulate engagement are now detectable by Google and Microsoft, and the gap between simulated and real warmup behavior is widening every quarter.

The mailbox-estate sizing math

Most founders working out volume planning start at the wrong end. They ask “how many emails do we need to send per week to hit pipeline goals” and try to back into mailbox count. The right calculation:

Step 1: target volume per day. If you need 1,000 cold sends/day, that’s your target.

Step 2: divide by safe per-mailbox cap. 1,000 / 35 = ~29 mailboxes needed at steady state.

Step 3: distribute across domains. Best practice is 2-3 mailboxes per domain, so 29 mailboxes = 10-15 domains. See multi-mailbox setup for per-domain provisioning.

Step 4: add buffer for warmup, rotation, and downtime. Real estate sizing is 1.3-1.5× the steady-state math. So 29 active mailboxes = 35-45 mailboxes provisioned, with the extras either in warmup, cooling down after a reputation event, or held in reserve.

Worked example for typical B2B teams:

The send-pacing rule — spread is as important as volume

30 sends in one hour and 30 sends spread over 8 hours are different signals to a recipient’s inbox provider, even though the daily count is identical. Bulk-sending tools and human typing produce very different temporal patterns; spam classifiers read this.

The operational rules:

• Pacing: spread sends across recipient business hours. For US-focused outbound, 8am-6pm recipient-local-time is the right window.
• Inter-send delay: a randomized 60-300 second gap between sends from the same mailbox is the typical config. This produces a human-like sending rhythm rather than a burst.
• No weekends: cold sends on Saturday/Sunday are a bot signal. Stop sends Friday afternoon, resume Monday morning.
• No overnight: 11pm-7am recipient-local-time sends are statistically associated with bulk senders and trigger filters.
• Volume ramp on Mondays: many teams send 60-80% of the week’s volume Tue-Thu. Mondays accumulate weekend backlog and overall inbox noise; Friday afternoons see lower open rates. Pacing isn’t flat across the week.

What happens when you push past the cap

The failure mode is gradual, not sudden. A mailbox sending 80-100 cold emails per day looks fine for the first week — opens and replies stay normal. The damage accrues in three stages:

• Week 1-2: inbox placement quietly drops. Reply rates fall 20-30%. The campaign tool still reports “sent” — but a growing share is landing in promotions or spam.
• Week 3-4: spam complaint rate climbs as more recipients see the message as a blast. Cumulative reputation damage starts to compound. Open rates fall.
• Week 5+: the mailbox is effectively burned. Inbox placement at major providers is 30-50%. Recovery requires 30-60 days of low-volume, high-engagement sending — or replacing the mailbox.

The asymmetry: a burned mailbox costs you 1-2 months of useful sending; a fresh mailbox costs $6-12/month on Google Workspace. The math says to err on the side of more mailboxes at lower per-mailbox volume. See why separate sending domains for the reputation-contagion case.

Per-provider quick reference

Where this fits

This chapter sits at the operational layer between multi-mailbox setup(how to provision the mailboxes) and the authentication/warmup chapters in the email-infrastructure cluster (what to do once they’re live). The sends-per-day cap is the single number that drives mailbox-estate sizing — every domain procurement decision flows from it.

The next question, after sizing the estate, is which subset of mailboxes to send a given campaign from. That’s campaign-orchestration territory, not infrastructure — but the per-mailbox cap defined here is what the orchestration logic schedules against.