How Many Cold Emails Per Day? The Safe Sending Math

Eight to ten cold emails per mailbox per day. That is the volume that keeps a warmed mailbox healthy over months rather than weeks. You can push toward fifteen when engagement is genuinely strong, and below five you are paying for infrastructure you are not using. Everything else in cold email volume planning is multiplication from that one number.

It is lower than every provider's published limit, and deliberately so. The published limits describe the point at which the provider stops you. The number above describes the point at which the provider starts distrusting you, and those two are weeks apart.

Because the published limits are abuse ceilings, not deliverability guidance. Google Workspace allows two thousand messages a day from a paid account, five hundred from a trial, and three thousand external recipients a day. Exchange Online allows ten thousand recipients a day, thirty messages a minute and five hundred recipients per message. Hit any of those and the provider throttles or suspends you. Nothing in those numbers says anything about whether your mail reaches an inbox.

Placement is decided somewhere else entirely, by a filter that profiles the mailbox: how much it sends, how suddenly that changed, how many recipients reply, archive, delete without opening or press the spam button. A brand-new mailbox that sends four hundred messages on a Tuesday is inside every published limit and has told the filter everything it needs to know.

Google publishes the one threshold that matters for this: keep the spam-complaint rate below 0.1 percent and never let it reach 0.3 percent. At eight sends a day a single complaint takes a mailbox past 0.1 percent for that day, which is a useful way to feel how thin the margin is. At four hundred sends a day you will not notice the complaints until placement has already gone.

The arithmetic runs in one direction: target volume, divided by per-mailbox volume, gives mailboxes; mailboxes divided by two gives domains; mailboxes divided by the tenant ceiling gives tenants or workspaces. Everything else follows.

For a hundred cold emails a day: ten mailboxes at ten a day, five domains, one tenant. That is a one-person fleet and it costs roughly a hundred dollars a month in seats and domains if you buy Google Workspace, or the price of a single server if you self-host.

For a thousand a day: a hundred mailboxes, fifty domains, four Microsoft tenants at twenty-five users each. This is the size where the operational work stops being a side task. Fifty domains means fifty sets of DNS records to write, verify and keep correct, and a steady trickle of replacements as individual domains age out.

For five thousand a day: five hundred mailboxes across about two hundred and fifty domains. At that scale the infrastructure is the business. DNS drift, expiring tokens, tenant limits and burned-domain replacement are a full-time job, and the cost of the fleet is smaller than the cost of running it by hand. This is the point where every serious sender either automates or hires.

Why not fifty emails each from twenty mailboxes instead? Because provider filters profile per-mailbox behaviour, and sudden volume from a small pool looks exactly like what it is. Many mailboxes at human volume beat few mailboxes at machine volume, every time, and the gap widens the longer both run.

A fresh mailbox starts at its provider's floor, a handful of messages a day, and ramps over two to three weeks. During the ramp, warm-up traffic and campaign traffic share the same daily cap, so a mailbox warming at eight a day has room for two campaign sends, not ten. Forcing week-three volume in week one is the most common self-inflicted deliverability wound there is, and it is invisible until it is expensive.

Plan capacity on steady-state numbers and let the ramp run. If the campaign has to start on a date, buy the mailboxes three weeks before that date. There is no version of this where the wait can be bought out: pre-warmed mailboxes from a vendor carry someone else's history, on a domain you usually do not own.

The eight-to-ten number is total outbound per mailbox per day, which means a sequence with follow-ups reaches far fewer new people than the raw number suggests. A three-step sequence in a steady state sends roughly one new-lead email for every three sends once the pipeline is full. Ten sends a day per mailbox is therefore about three new prospects a day per mailbox, or sixty-six a month.

That arithmetic is the one most plans get wrong. A team that needs to reach three thousand new people a month does not need a thousand sends a day; it needs roughly three thousand sends a day, which is fifteen times the mailbox count they budgeted. Decide the new-lead rate first, multiply by the number of steps in the sequence, and size the fleet from that.

It also argues for shorter sequences than most templates suggest. Steps four and five of a sequence consume the same daily cap as step one and convert at a fraction of the rate. Cutting a five-step sequence to three does not reduce pipeline by two fifths; it usually increases it, because the freed capacity goes to new people.

Four signals override the plan. A spam-complaint rate approaching 0.1 percent, which Google publishes as the line and 0.3 percent as the point of real trouble. A hard-bounce rate above about two percent, which reads as a bought or stale list and is the fastest way to lose a new domain. A reply rate that falls while volume is flat, which usually means placement moved to Promotions or junk before anyone told you. And provider-level refusals, the 5.7.x family, which are a verdict about your sending rather than about the recipient.

When any of those appear, the correct response is to reduce volume on the affected mailboxes and domains, not to rewrite the email. Content is the last variable to look at, because it is the one filters weigh least once a domain has a history. A perfect email from a domain that has been complained about still lands in spam.

Check Google Postmaster Tools for the domains you send to Gmail from. It is free, it shows the complaint rate the filter actually sees, and it is the only place you will see your reputation described in Google's own words rather than inferred from your open rates.

Gmail is the most forgiving of volume and the least forgiving of complaints. It profiles the mailbox rather than the domain, recovers reasonably well after a bad week, and publishes what it sees through Postmaster Tools. If your recipients are mostly at Gmail, the eight-to-ten discipline plus a clean list is usually enough.

Microsoft 365 is the opposite. It weighs the domain and the sending IP more heavily, it reacts slowly in both directions, and it tells you almost nothing. A Microsoft-heavy recipient list rewards patience: longer warm-up, lower volume for the first month, and a tolerance for the fact that you will not know your standing until replies tell you.

Corporate gateways in front of either, the secure email gateways that large companies still run, behave differently again. They keep keyword weights, they rate-limit by connection, and they will quietly quarantine a sender who arrives with perfect authentication and a promotional vocabulary. That is the one place where the words in the email still move the outcome, and it is why a spam-words pass is worth the five minutes even in 2026.