How Many Cold Emails Should You Send Per Day

Most operators treat cold email volume like a dial they can turn up. We have sent over 8 million hyper-personalized cold emails this year across 50+ B2B campaigns, and volume is not a dial, it is a ceiling that your infrastructure sets for you. Below, the per mailbox ceiling with the provider documentation behind it, the sizing math that runs backwards from your target, and the week by week ramp curve.

How Many Cold Emails Can You Send Per Day Per Mailbox?

The question gets asked as though there is a single sending point. There is not. Every working cold email setup in 2026 runs a pool: several domains, a handful of mailboxes on each one, a warmup layer underneath all of them, and a sending tool rotating traffic across the whole pool.

So the binding constraint is never emails per day. It is emails per mailbox per day, times the number of mailboxes you have warmed and are willing to risk. Push 150 a day through one mailbox and you get filtered inside a fortnight. Spread 480 a day across 16 mailboxes on 6 domains and placement holds.

Per Mailbox Sending Ceiling

The highest number of outbound messages a single mailbox can send per day without the provider reading it as bulk behavior. In 2026 that ceiling sits at 20 to 40 messages a day per mailbox for cold outreach, warmup included. It is a reputation limit, not a technical one, which is why it sits so far under the documented provider cap.

Two numbers get confused constantly, so it is worth separating them before any of the math makes sense. The documented limit is where the provider stops accepting mail. The safe ceiling is where the provider stops trusting you. The first is published. The second is inferred from behavior, and it is the one that decides whether your campaign works.

The last row is the one people argue with. It is also the one that costs the most when it goes wrong, because the primary domain is where your invoices, contracts, and support mail live. Our breakdown of what a secondary domain is and why the primary never sends covers the separation in full.

What Are the Real Provider Limits, and Why Is the Safe Number So Much Lower?

Start with what the providers actually publish, because almost nobody writing about this reads the source.

Google documents a limit of 2,000 messages a day per Workspace user, 3,000 total recipients a day, and 2,000 external recipients a day in its Gmail sending limits for Workspace accounts. A trial account is capped at 500. Microsoft documents a recipient rate limit of 10,000 recipients a day on a 24 hour sliding window and a message rate limit of 30 messages a minute in the Exchange Online service limits. Microsoft then layered an External Recipient Rate Limit of 2,000 external recipients in 24 hours on top of that, which matters a great deal for outbound because every cold recipient is external by definition.

None of those numbers are permission slips. They are the point where mail stops leaving the building. Reputation filtering starts an order of magnitude earlier.

The second set of rules is the bulk sender layer. Google's email sender guidelines classify anyone sending 5,000 or more messages a day to personal Gmail accounts as a bulk sender, and bulk senders have to pass SPF and DKIM, align DMARC, honor one click unsubscribe per RFC 8058, and hold spam complaint rates under 0.3 percent. Google's own guidance is to stay under 0.1 percent. Above 0.3 percent, mail starts getting rejected at the SMTP layer rather than quietly filed in spam, which is a harder failure and a slower recovery.

Now the gap. MailReach's sending volume analysis puts the practical safe band at 30 to 50 a day per mailbox. Lemlist's infrastructure guidance caps a ramped mailbox near 40 a day, ages every new domain at least 4 weeks before full volume, and holds roughly 10 percent of the pool in reserve. ScaledMail's volume guide runs more conservative still at 15 to 25 a day per mailbox on Google Workspace. Every credible source lands somewhere between 15 and 50, which is 2 percent to 5 percent of the documented Google limit.

We run the band at 20 to 40 and sit nearer 30 on anything new. The reason is asymmetry. Sending 25 instead of 40 costs you a few more mailboxes. Sending 80 instead of 40 costs you the domain, and recovering a burned domain takes 4 to 8 weeks of near zero output. The cheap mistake and the expensive mistake are not the same size, so bias toward the cheap one.

How Do You Size Infrastructure Backwards From a Target Volume?

This is the part almost every guide skips. People buy domains first and work out capacity afterwards, which is how you end up with 8 domains for a 2,000 a month campaign and 2 domains for a 20,000 a month campaign. Run it the other way round. Start from the number of sends the campaign needs, then solve for mailboxes, then for domains.

The arithmetic is 3 steps:

1. Daily sends. Monthly target divided by 21 working days. Weekends are out, because a cold email landing on a Saturday reads as a machine and gets buried by Monday.
2. Mailboxes. Daily sends divided by 30. Use 30 rather than 40 so the ceiling has headroom when one mailbox needs to be pulled.
3. Domains. Mailboxes divided by 3. More than 3 or 4 mailboxes on a single domain makes the domain itself look like a sending appliance, even when each mailbox behaves.

Read the 10,000 row carefully, because that is the row most B2B programs actually live on. 10,000 sends a month is 16 mailboxes across 6 domains. Not 2 domains worked hard. Not 30 domains bought in a panic. Six, each carrying under 100 sends a day, each with its own authentication records and its own reputation to lose independently of the others.

Then add the reserve. Build 10 percent more capacity than the target needs, so that pulling one domain offline for a week does not put you under plan. Our multi domain sending strategy covers the ratio math and the retirement rules, and setting up email domains for outbound covers the DNS order of operations so the records are live before the first warmup message leaves.

One sizing mistake worth naming: people size for sends and forget to size the list. 10,000 sends a month against a 4 step sequence is 2,500 new contacts a month, every month, all verified. If list supply cannot keep up, the sequence starts recycling contacts across different mailboxes, which reads worse to a recipient than never emailing them. Building a cold email list from scratch walks the supply chain, and email verification is what keeps the bounce rate from eating the ramp you just paid 5 weeks for.

What Does the Ramp Curve Look Like, Week by Week?

The ramp is where most programs die, and they die in week 1. Teams buy 6 domains, spin up 16 mailboxes, and start sending 480 a day on the Monday. Two weeks later half the pool is in spam and they are buying replacements. That is the single most expensive way to run outbound.

The reason is velocity. A domain that goes from 0 to 480 a day inside 48 hours has told every receiving server exactly what it is. A domain that walks to 480 over 5 weeks looks like a business hiring salespeople. Same destination, completely different signal. Woodpecker's warmup guide and most other practitioner sources converge on the same rule: never raise volume more than about 20 percent in a week.

Three things in that table are load bearing.

First, week 1 sends no cold email at all. The mailbox is building a baseline of normal correspondence so the cold traffic in week 2 arrives on top of something. Warmup explained covers the signals providers read, and warming up a new email domain covers the ramp specific to a domain nobody has ever seen before.

Second, warmup counts inside the ceiling, and it tapers. A mailbox running 25 warmup sends and 35 cold sends is a 60 a day mailbox as far as Gmail is concerned, because Gmail sees one outbound stream and cannot tell the two apart. Warmup never switches off entirely though. Cut it to zero and placement degrades within 2 to 3 weeks because the only thing left in the stream is cold.

Third, each column is a gate, not a schedule. If bounce rate is at 3 percent in week 2, week 3 does not happen. You hold volume, fix the list, and move when the number moves. Bounce rate causes and fixes is the diagnosis path, and SPF, DKIM and DMARC plus the DNS record layer cover the authentication failures that masquerade as content problems.

What Happens When You Send Too Many, and When You Send Too Few?

Both ends cost you, in different currencies. Too much volume costs infrastructure. Too little costs time, which is usually the more expensive of the two.

Too many per mailbox:

• You cross into bulk sender territory, which brings the 0.3 percent complaint ceiling and SMTP level rejections rather than soft filtering.
• Open rate falls first, then reply rate. The mail is landing in promotions or spam, so the copy never gets judged. Keeping cold email out of spam covers the fixes, and the measurement loop covers how you see it happening.
• Domain reputation damage is slow to undo. Google Postmaster Tools shows the reputation grade dropping days after the behavior that caused it, and recovery takes 4 to 8 weeks of reduced volume even after the cause is gone.
• Every domain in the pool is at risk, not just the one that broke, because a sending tool rotating across a pool shares the list, the copy, and often the IP neighborhood.

Too few per day:

• You cannot test anything. 50 sends a day is 1,050 a month, which is roughly 1 meaningful split test per quarter. 500 a day is 10,500 a month and you learn 10 times faster on angles, subject lines, and segments.
• The warmup ratio inverts. A mailbox sending 20 warmup and 5 cold is 80 percent synthetic traffic, and providers eventually notice that the warmup pattern and the real pattern do not match.
• Time to the first booked conversation stretches out. At 50 sends a day and a 4.6 percent reply rate you get 2 replies a day. At 500 you get 23. Only one of those fills a calendar.

Which Numbers Tell You to Scale Up or Pull Back?

Volume decisions should come off a dashboard, not off a feeling about the quarter. These are the 5 numbers we gate every volume increase on.

Read that table left to right before every increase. Any single column in the pull back band stops the increase, and 2 columns in the hold band stops it too.

The ordering matters as much as the thresholds. Open rate is the leading indicator and reply rate is the lagging one. By the time reply rate moves, the domain has been degrading for 2 or 3 weeks and needs weeks more to come back. Watch the leading indicator weekly, not the lagging one monthly. Deliverability monitoring covers the full metric stack and the alert thresholds, and inbox placement testing covers how to read a seed report and where seed lists quietly lie to you. Tools like GlockApps and Google Postmaster Tools see different halves of the picture, so run both.

One pattern shows up in nearly every campaign we inherit: the team scaled too fast and pulled back too late. The early signal is subtle, open rate sliding from 55 percent to 46 percent over 3 weeks. Nobody reacts to a 9 point slide. Then reply rate halves and everybody reacts at once, on a domain that is already 6 weeks from healthy. The fix is a weekly ritual, not a bigger alarm. Deliverability as a measured system is the hub for that whole stack.

Why Does Changing the Ask Change the Volume Math?

Everything above assumes the ask is a pitch. Change the ask and the arithmetic moves, because volume only ever multiplies the reply rate you already have.

Instantly's 2026 benchmark report puts the platform wide average reply rate at 3.43 percent, with top quartile senders above 5.5 percent, drawn from hundreds of thousands of businesses. Growth Engineer's analysis of 14 million sends lands in the same territory. Across our book we sit at 4.6 percent, and we have handled over 95,000 positive replies this year. That gap did not come from sending more. It came from changing what the email asks for.

An invite to be interviewed on a show is a different ask than a meeting request. It flatters instead of taking, it needs no case study, and it survives being read by someone who has never heard of you. Invite versus pitch is the full argument, and podcast invite reply rate benchmarks carry the segment level numbers.

The practical consequence is that the volume target drops. A program that needs 30,000 pitch sends a month to fill a calendar might need 10,000 invite sends to fill the same one, which is 16 mailboxes instead of 48 and 6 domains instead of 16. That is a smaller infrastructure bill, a shorter ramp, and far less reputation exposed. Our own client programs run on 10,000 invites a month precisely because the ask carries more of the work than the volume does.

The ceiling per mailbox does not move though. Invites obey the same 20 to 40, the same warmup taper, the same complaint thresholds. Domains and warmup for podcast invites covers the invite specific ramp, and how many sending domains an invite program needs runs the same sizing math on invite volume. What changes is how far each send gets you, which is the only lever that actually compounds. Defining the ICP as a deliverability constraint and personalizing at scale are the two other levers that move reply rate without touching the ceiling.

That is also why the guarantee we offer is written in conversations rather than sends. 30 recorded conversations with your ideal buyers in 90 days, or your money back. Nobody buys sends. Editing and publishing every episode is included, invites go out over email only, and the client owns every recording on their own show.

Volume Is a Constraint, Not a Strategy

The programs that win are not the loudest ones. Across 50+ campaigns we run, the highest performing are rarely the highest volume. They are the ones with the tightest list, the strongest ask, and a per mailbox ceiling nobody is tempted to cheat.

Treat the ceiling as fixed and the question gets better. Given 30 sends per mailbox per day, what has to be true about the list, the ask, and the follow up for each one to be worth sending? Reply rate benchmarks tell you whether your answer is working, positive reply rate tells you whether the replies are the right kind, and the follow up sequence decides how much of the list you reach twice without buying a single extra mailbox.

The goal is not to send as many cold emails as your provider allows. It is to send as many strong ones as your infrastructure can carry without the providers deciding what you are. Those are 2 very different targets, and only one of them is still working in 12 months.

Volume gets you at bats. The ask, the list, and the ramp decide whether those at bats convert. Build the pool to the number, walk it up over 5 weeks, watch open rate weekly, and spend the rest of your attention on the 2 inches of copy that nobody can buy their way out of.