How much cold email can one company actually send in 2026

A ten-person company running cold outbound as well as it can be run tops out around 15,000 to 20,000 cold emails a month in 2026. That is the ceiling. Not the ceiling for a lazy program, the ceiling for a disciplined one: verified lists, enough domains, warming that never stops, and mailboxes held to their real daily limit. Cold email capacity is a multiplication problem, per-mailbox daily volume times mailbox count times sending days times uptime, and every factor in that chain is contested. This piece puts numbers on each one and shows where the product stops growing.

Both of the popular answers are wrong. The 100-a-day-per-mailbox numbers still circulating in older guides describe a filtering era that ended; the sub-20 numbers are excess caution, priced for senders who skip verification. And the factor that actually kills programs is not the per-mailbox number at all. It is the one nobody budgets for: domain count.

The real number per mailbox

A well-warmed mailbox on a healthy domain, sending to a verified list, holds 30 to 50 cold emails a day. That is the working band across the email deliverability tools we track, and it has been narrowing from both directions.

From above: the 100-a-day figure came from a period when mailbox providers scored senders mostly on complaint rate and authentication. Behavioral scoring changed that. A mailbox that sends 100 near-identical messages a day to strangers, gets few replies, and generates no ordinary correspondence does not look like a person, whatever its SPF record says. Push a mailbox to those volumes today and placement decays within weeks, not months, and it decays quietly, which is the expensive part.

From below: infrastructure vendors increasingly publish guidance near 15 cold emails a day per mailbox, Maildoso among them, and that number is a warranty figure. It protects a mailbox whose owner uploads an unverified list, skips the ramp, and sends the same template five days running. Price in the discipline, a verifier in front of every import, warming behind every mailbox, and rotation across a real domain footprint, and the safe band moves up to 30 to 50. Our own under-$150 build-out, the companion piece for anyone assembling this stack from zero, budgets below that band deliberately, because a brand-new operation in its first months has not earned the top of it yet.

The verification discipline is what separates the bands, and it is nearly free. MillionVerifier charges $39 for 10,000 checks, charges only for good and bad results, and lets unresolved catch-alls through at no cost. Four tenths of a cent per address is the toll for sending 50 a day instead of 15. Every one of the 17 email verification tools we track offers a free plan, which tells you what the category thinks the excuse for skipping it is worth.

Domains are the real denominator

Here is the mistake that actually ends programs. A team provisions ten mailboxes, puts them on one or two domains because domains feel like overhead, and runs happily for a quarter. Then one campaign misjudges its audience, or one bad import slips a bounce spike through, and the whole operation goes dark at once. Not one mailbox. All of them.

Reputation damage lands at the domain level. Providers score the sending domain first and the address second, so ten mailboxes on one domain are not ten independent assets; they are one asset with ten doors. The unit of failure is the domain, which means the unit of capacity planning has to be the domain too. Mailbox count is how big your engine is. Domain count is how many engines you have.

The stable shape is two to three mailboxes per domain, and the domains themselves are the cheapest line in the entire stack. Mailforge registers a .com for about $14 a year and sells mailbox slots at $3 a month on annual billing; Primeforge provisions real Google Workspace and Microsoft 365 mailboxes at $4.50 a slot; Zapmail bundles ten Google mailboxes for $39 a month. Both Forge products treat slots as capacity rather than named mailboxes, so retiring a burned address and minting a replacement costs nothing, which is exactly the right model for infrastructure you expect to lose pieces of. Eight to ten domains instead of two adds roughly $10 a month in registration, a rounding error against what a domain-level failure costs.

Warming is not a phase, it is a state

The most persistent bad mental model in cold email is that warming is a project with an end date. Connect the mailbox, run the warm-up tool for two to four weeks, graduate, cancel. That description mistakes the ramp for the whole job.

What actually ends after two to four weeks is the ramp: the period where a new mailbox builds volume from nothing to its working level. What does not end is the engagement floor underneath it. Mailbox providers score senders on recent behavior, and recency is the operative word. A mailbox whose entire recent output is cold email to strangers, with reply rates in the low single digits and no threads that deepen, drifts back toward the profile of a spammer no matter how good its history was. Sender reputation is not a qualification you earn. It is a balance that decays. The moment the only mail a mailbox sends is cold, the balance starts draining, and "finished warming" just means the drain has not shown up in placement yet.

So the correct steady state is cold volume and positive engagement running side by side, permanently. This is where MailStrike is the clearest worked example in the category, because its architecture takes the never-finished position literally. Each connected mailbox gets its own AI Persona, a synthetic correspondent with a distinct writing style, schedule, and reply habits, and the Persona does not retire when the ramp ends: it keeps holding conversations 24/7 alongside live campaigns, so the mailbox's outbound mix never becomes purely cold. If reputation dips, the Persona pauses protectively rather than sending into the damage. The monitoring layer runs on the same always-on assumption, with reputation scores refreshing every 15 minutes and domains checked daily against more than 100 blacklists, because a decay process needs continuous measurement, not a monthly check-in.

The same logic extends to the send itself. MailStrike's Seedlist add-on lets a sender push the actual campaign into thousands of trusted Gmail and Outlook inboxes before launch, collecting per-provider placement feedback and a wave of positive engagement on the exact content, and its built-in placement test sends from the user's own connected account with results in about three minutes. The principle both features encode is the one this whole section argues: engagement testing belongs into or before a send, not after one fails. Pricing makes the always-on posture affordable at fleet scale, $29 a month covering up to 10 mailboxes and $79 covering up to 50 on flat tiers, so warming 30 mailboxes forever costs less than most single per-mailbox subscriptions. Budget warming as rent, not as a construction cost.

What a burned domain is actually worth

Domains die anyway. A content mistake, an unlucky complaint cluster, a bounce spike from a list that aged faster than expected, and placement on one domain collapses. The standard response is to rest it for two weeks and resume, and the standard response is theater. Two weeks is long enough to feel responsible and short enough to change nothing; the reputation that took a quarter to damage does not repair on a sprint cadence.

The playbook that works is longer and cheaper than it sounds. Pull the domain out of rotation for 60 to 90 days. Keep light warming engagement on it so the domain is not simply silent, then rewarm it from zero exactly as if it were new, full ramp, no shortcuts, before it carries cold volume again. Treated this way, most burned domains come back usable. The 60 to 90 day rest is why the 8 to 10 domain footprint matters twice: it caps the blast radius on the way down and it means losing one domain costs you a tenth of your capacity for a quarter instead of all of it.

A hard blacklist listing changes the order of operations. Resting does not remove a listing; time plus silence is not a delisting process. Check the major lists first, follow the delisting procedure for any hit, and only then start the rest-and-rewarm clock. This is a place where monitoring earns its subscription, since MailStrike's daily checks against Spamhaus, SORBS, Barracuda and the rest of its 100-plus list sweep, with delisting guides attached, turn a listing from a mystery you diagnose in month two into an alert you action on day one.

The ceiling, worked backward

Put the factors together and the ceiling assembles itself.

Factor	Realistic value	Running arithmetic
Monthly cold sends, the ceiling	15,000 to 20,000	the number everything below must produce
Sending days per month	About 21	cold email performs on weekdays
Required daily fleet output	700 to 950	ceiling divided by sending days
Per-mailbox daily volume	30 to 50	the 2026 band for a warmed mailbox
Mailboxes needed on paper	20 to 30	daily output divided by per-mailbox volume
Mailboxes per domain	2 to 3	the blast-radius limit
Domains required	8 to 10	fleet divided by density
Fleet uptime	80 to 85 percent	ramps, protective pauses, one domain resting

The uptime row is the one honest programs respect and optimistic spreadsheets omit. At any given moment some slice of the fleet is not at full output: a replacement mailbox is three weeks into its ramp, a Persona has paused a wobbling address, one domain is serving its 60 days. Run the middle of every band, 25 mailboxes at 40 a day across 21 days at 85 percent uptime, and you get roughly 17,800 sends a month. That is the whole result. A ten-person company doing everything right lands between 15,000 and 20,000, and the operators who claim multiples of that are either burning domains on a schedule or counting sends their prospects never saw.

The full stack at that ceiling stays cheap in software terms: about $90 a month in mailbox slots, $12 in domains, $79 for warming and monitoring across the fleet, a sequencer such as Instantly at $37 or Smartlead at $39 (both meter contacts and volume rather than mailboxes, so the fleet connects at no extra cost), plus verification. Call it $250 to $300 a month, all in.

Past the ceiling, the money should change jobs

Here is the punchline the capacity math has been building toward. You can buy past the ceiling. Nothing stops a company from running 20 domains and 60 mailboxes, and the cost curve is almost perfectly linear: twice the domains, twice the slots, the next warming tier, twice the verification volume. Capacity scales at retail price.

Reply quality does not. The first 15,000 sends a month go to the best-fit accounts a ten-person company knows how to name. The next 15,000 go to the second-best list, written by the same two people, now stretched across twice the surface, into markets that have already seen the first wave. Cost per send holds flat while value per send falls, and somewhere just past the ceiling the lines cross. At that point the marginal outbound dollar buys less than the same dollar spent on ads or content, channels that compound with volume instead of diluting. The ceiling is not a limitation to engineer around. It is the signal that this channel is fully spent and the budget should diversify.

The short version

A warmed mailbox with a verified list holds 30 to 50 cold sends a day in 2026; the old 100-a-day figures are dead and the 15-a-day figures are caution priced for people who skip verification. Plan capacity by domains, not mailboxes, at two to three mailboxes per domain, because reputation fails at the domain level. Warming never finishes, so budget continuous engagement alongside cold volume, and rest a burned domain 60 to 90 days with a full rewarm, not two performative weeks. Do all of it right and a ten-person company tops out near 15,000 to 20,000 sends a month for roughly $300 in software. Past that line, spend the next dollar somewhere that compounds.