How many domains and inboxes you need for cold email in 2026

You bought four domains because a video said four. Twelve mailboxes on Google Workspace, three per domain, EUR 82 a month on the invoice. Two weeks of warmup, then 40 emails a day out of each one, and by week three the open rate had fallen through the floor and a test send to your own Gmail landed in spam. So you bought four more domains, which is the move the next video recommends, and the number got worse.

The sizing question is the one nobody answers with arithmetic. Vendor blogs hand out a per-inbox number, and those numbers disagree by more than 16x depending on who is selling what. The number you need is not a rule of thumb. It falls out of one calculation that starts at the meetings you want, and the rest of this page runs that calculation.

How many domains and inboxes do you actually need?

Enough inboxes to carry your daily send volume at 25 to 50 cold emails each, spread across domains at no more than 3 to 5 inboxes per domain. For a founder-led agency going after eight meetings a month, that lands at 17 to 27 inboxes on 6 to 9 sending domains, and the spread in that range is your reply rate rather than your budget.

Two ratios do all the work, and both of them are about containment rather than throughput. The per-inbox cap exists because a single mailbox sending more than about 50 cold messages a day looks mechanical to a filter. The per-domain cap exists because reputation attaches to the domain, so one inbox collecting complaints takes its neighbours down with it. Everything else is multiplication.

What makes the question hard is that it gets asked in the wrong direction. A founder asks how many inboxes they need, as though the inbox count were the input. It is the fifth step of a chain whose first step is a meeting target, and the steps in between are reply rate, sequence length and working days. Start at the wrong end and you are guessing with precision.

Why does every guide give you a different number?

Because almost every guide on this question is published by a company selling either inboxes or the software that sends from them, and the recommended number moves with the business model.

Line the current advice up and the spread is not subtle.

Source	Cold sends per inbox per day	Inboxes per domain	Warmup before full volume
Instantly, strategy docs	30 to 50	max 3 to 5	14 days, 21 for AirMail
Instantly AirMail, enforced cap	50 hard maximum	5 maximum	3 weeks
Unify, September 2026	25 default, up to 65	across 2 to 3 or more	about 3 weeks
ScaledMail, July 2026	15 to 25 Google, 10 to 25 Microsoft	2, up to 3 on Microsoft	not stated
Growth Engineer, May 2026	20 to 50 aged, 0 to 5 in week one	4 to 10 across 2 to 3 domains	30 days or more
Puzzle Inbox, March 2026	15 Google, 3 Microsoft	3 maximum	14 to 21 days

Read the bottom row against the second one. Puzzle Inbox's formula puts a single Microsoft inbox at 3 cold sends a day; Instantly enforces 50 on its own accounts. That is a 16x gap on the single most load-bearing number in the whole setup, and the consequence is concrete: at 500 sends a day, Puzzle Inbox's math produces 87 inboxes across 30 domains, while planning at 25 a day produces 20 inboxes across 7 domains. Same volume, four times the infrastructure, four times the warmup to supervise.

Neither number is dishonest. They are measuring different things. A per-inbox provider has no incentive to tell you that fewer inboxes would do, and a sending platform that bills by volume has no incentive to tell you to slow down. The tell is whether a source publishes the conditions under which its number holds, and most of them do not.

So treat every published figure as a starting prior and replace it with your own data at 60 days. lemlist makes the methodology point directly about reply rates, noting that published benchmarks "vary by as much as 7.6x depending on methodology," with Saleshandy's 53 million email analysis landing at 3.7 percent against Belkins' stricter 7.5 million email study at 0.45 percent. The same warning applies to everything else on this page.

Is the real limit the provider's cap or your sender reputation?

Reputation, by a factor of 40 to 400. The published platform limits are so far above the safe cold-sending volume that they never bind, which is why reading your provider's terms tells you nothing useful about how many inboxes to buy.

The primary numbers are easy to check. Google's documented Gmail sending limits for a paid Workspace user are 2,000 messages a day, 10,000 total recipients a day and 3,000 external recipients a day. Microsoft's Exchange Online limits are a recipient rate limit of 10,000 recipients per day on a 24 hour sliding window and a message rate limit of 30 messages per minute. At 30 a minute you could clear a day's safe cold volume in under two minutes.

What actually binds is complaint rate. Google's sender guidelines require all senders to keep spam rates reported in Postmaster Tools below 0.3 percent, and recommend staying below 0.10 percent while avoiding ever reaching 0.30 percent. Senders above 5,000 messages a day to Gmail also need SPF and DKIM both passing, DMARC set up, and one-click unsubscribe on marketing mail. At 0.3 percent, three complaints out of a thousand sends is the enforcement line, which an agency sending 670 a day can trip with two annoyed recipients on a bad morning.

Microsoft is blunter about the underlying posture. Its outbound spam documentation states that using Microsoft 365 to send bulk email "isn't a supported use of the service, and is permitted only on a best-effort basis," that accounts exceeding undisclosed spam thresholds get blocked from sending, and that senders should avoid their primary domain and configure dedicated subdomains with SPF, DKIM and DMARC. The company also throttles outbound mail from default onmicrosoft.com domains to 100 external recipients per organisation per 24 hours. None of those are published volume caps. They are reputation controls, and they are the thing your architecture has to respect.

How do you size it from your meeting target?

Run six steps in order, starting at the number of meetings you want on the calendar and ending at a domain count. Every step is a division, and every input is either a benchmark or a decision you already made.

Step one is the target. Pick meetings a month, not leads and not sends, because meetings are the only number in the chain that connects to revenue. Step two converts meetings into contacts, and this is where the arithmetic gets interesting, because two independent routes land in the same place. Multiply Instantly's 3.43 percent average reply rate by the roughly 14 percent of replies that are positive and the 40 percent of positive replies that become a meeting, and you get 0.19 percent, or about 520 contacts per meeting. Agentic Demand's benchmark set puts the meeting-booked rate per email sent at 0.1 to 0.5 percent for the average campaign. Two chains, built from different data, agreeing inside the same band.

Step three turns contacts into sends. A four-step sequence does not send four emails to everyone, because replies, bounces and unsubscribes pull people out early, so 3.5 per contact is a fair working figure. Step four divides by 22 working days to get a daily volume. Step five divides that by 25 to get inboxes, and step six divides inboxes by 3 to get sending domains.

Two warnings about using this. The 0.19 percent figure describes average performance across a platform full of campaigns, including bad ones, so a well-targeted list to a named buyer beats it comfortably. And warmup traffic competes with campaign traffic inside the same daily cap, a point ScaledMail makes well when it notes that warmup "compresses your cold sending room." Their formula adds a 1.2 buffer for exactly that reason, which is a sensible thing to copy.

What do three performance levels cost in inboxes?

The same eight meetings a month needs 60 inboxes at a weak reply rate and 17 at a strong one, so your offer sets your infrastructure bill before your budget does.

Step	Weak offer, 1.5% reply	Average, 3.43% reply	Strong offer, 5.5% reply
Meetings wanted a month	8	8	8
Contacts to reach	about 9,500	about 4,200	about 2,600
Emails sent a month	33,000	14,700	9,100
Sends a working day	1,500	670	415
Inboxes at 25 a day	60	27	17
Sending domains at 3 each	20	9	6
Mailboxes and domains a month	about $205	about $90	about $60

The cost row uses Mailforge's $3 per mailbox a month and $14 a year per .com, and it excludes the sending platform, which adds $39 on Smartlead's Base plan or $194 on Instantly's Scale bundle. The middle column is the one to plan against, because 3.43 percent is the measured average rather than an aspiration.

Now read the first and last columns against each other. Going from a 1.5 percent reply rate to 5.5 percent cuts the inbox count by 72 percent, the domain count by 70 percent, and the warmup you have to supervise from 60 mailboxes to 17. It also cuts the volume of strangers you annoy by three quarters, which is the part that protects the domains you keep. Two weeks spent rewriting the offer and tightening the buyer definition does more for the infrastructure bill than any purchase, and it costs nothing but attention. The guide on why an agency offer fails on cold outreach covers what that rewrite involves.

The table also holds a conservative assumption worth naming. It moves the reply rate and keeps the positive-reply share and the meeting conversion fixed, which is not what happens in practice. A sharper offer lifts all three, so the right-hand column understates the gain rather than overstating it.

Why can you not send cold email from your main domain?

Because the domain your clients reply to is the one asset in this setup you cannot replace, and cold sending is the activity most likely to damage it. Microsoft's own guidance is to avoid using addresses in your primary email domain as senders for bulk mail, since doing so "can affect the delivery of regular email from senders in the domain."

Reputation is domain-level, and it is sticky. Growth Engineer's rebuild plan for a burned domain runs 90 days and the piece explains why compression below that is not possible: mailbox providers hold negative reputation signals for about that long. So a domain that collects complaints in March is still carrying them in June. If that domain is the one on your invoices, your proposals and your client threads, you have turned a campaign problem into a business problem.

The fix is cheap and slightly unglamorous. Register close variants of your brand, the way Instantly's strategy docs describe with get and try prefixes, point them at your real website with a redirect, and treat each one as consumable. Cloudflare Registrar sells at registry cost with no markup, and Mailforge lists .com registrations at $14 a year, so nine sending domains is roughly $126 a year. That is the cheapest insurance in the entire stack, and it also lets your main domain keep a strict DMARC policy while the sending domains sit at p=none until you know what passes.

How many inboxes go on one domain, and which domains do you buy?

Three inboxes per domain as the default, five as the hard ceiling, and the domains are short brand variants on .com wherever you can get them. The number is about containment, so the question to ask is how much volume you are willing to lose when one domain goes bad.

Work the blast radius. At 3 inboxes per domain and 25 sends each, a burned domain costs you 75 sends a day. At 5 per domain it costs 125. Spread the same 27 inboxes over 9 domains and any single failure takes out 11 percent of your capacity, which you can absorb while a replacement warms. Put them on 5 domains and a failure takes out a fifth. Instantly caps AirMail at 5 accounts per domain and ScaledMail argues for 2, going to 3 on Microsoft; the whole published range sits between 2 and 5, so 3 is the defensible middle.

On the domains themselves, a few rules earn their place. Prefer .com, because Instantly's docs note that .agency and .tech also perform while .com remains the baseline, and a filter has seen fewer bad actors on it. Keep the name legible and close to your brand so a recipient who checks does not find a string of characters. Redirect every one of them to your real site, because an empty domain is itself a signal. And buy them in one batch, in a single registrar account, so renewals and DNS do not scatter.

Name the inboxes after real people on your team, with addresses that match how your company normally writes them. A sending domain with firstname@ addresses behaving like a small company is the shape you want. Three mailboxes called info@, sales@ and hello@ on the same domain is the shape a filter was built to catch.

How long does warmup take, and what does the wait cost?

Fourteen days at the floor, three weeks for most providers, and four to six weeks if you want the ramp done properly. That makes warmup the longest lead time in the build, which is why domains get bought before the list is finished rather than after.

The vendor guidance converges here more than it does on volume. Instantly's warmup documentation asks for at least two weeks on a standard account and three on an AirMail account, about 10 warmup emails a day, and a health score above 90 percent before the first campaign goes out. Its 2026 benchmark report recommends starting at 5 to 10 sends a day and building over 4 to 6 weeks, with bounce rate held under 2 percent. Unify publishes a week-by-week ramp: about 5 a day in week one, 5 to 10 in week two, 10 to 20 in week three, then the 25 a day default. Growth Engineer wants 30 days or more and keeps warmup running in parallel with production sends on aged domains.

The cost of the wait is the part people underestimate. Three weeks of warmup on 27 mailboxes is three weeks of paying for mailboxes, domains and a sending platform while booking nothing, which at the average column is roughly $90 plus the sender for a month of no output. The expensive version is discovering in week four that the offer does not land, because you spent the warmup window building a list instead of testing a message. Run those two in parallel: domains warming, offer and buyer definition getting written, list being built. The full build sequence across all five jobs sets out the order.

Resist the two shortcuts. Turning warmup off once campaigns start removes the engagement signal that keeps an inbox healthy, and Instantly's own instruction is to never disable it. Jumping a new mailbox straight to 40 a day because the health score looks fine on day eight is the single most common way a founder burns a fresh domain, and the damage takes 90 days to undo.

Which inbox provider should you buy from?

An outbound-specialised inbox provider once you are past about ten mailboxes, and Google Workspace or Microsoft 365 below that. The setup that holds up at volume is specialist inboxes with DNS automated, because manually configuring SPF, DKIM and DMARC on nine domains is a morning you will not enjoy twice.

Provider	List price	What you get	Pick it when
Google Workspace	EUR 6.80 per user a month, EUR 5.44 on annual commitment	a real tenant, 2,000 messages a day on paper, highest sender trust	you run under 10 mailboxes and want the most trusted sender type
Microsoft 365 Business Basic	about $7 per user a month on annual commitment	Exchange Online, 10,000 recipients a day on paper, bulk sending unsupported	your buyers sit on Microsoft and you accept the lower safe cap
Instantly AirMail	$4 per account a month, $15 per domain a year	DNS done for you, warmup included, 5 accounts per domain, 50 a day hard cap	you already send in Instantly and want the setup to disappear
Instantly pre-warmed accounts	$10 per account a month, 5 account minimum	day-one sending at 20 to 25 a day, Google accounts only	you have to start this week and do not need to own the domain
Mailforge	$3 per mailbox a month on an annual plan, $14 a year per .com, 10 slot minimum	automated SPF, DKIM and DMARC, shared IP pool	you are running 10 or more mailboxes and want the cheapest compliant path
The wider per-inbox market	$1.24 to $4.50 per mailbox a month across 11 providers	varies; shared IP, instant provisioning	you are above 50 mailboxes and price per mailbox starts to dominate

Three shapes cover almost every agency. The lean shape is 6 mailboxes on 2 domains, Google Workspace, email only, which suits a founder testing whether cold outreach works for their offer at all and costs about EUR 41 a month plus a sender. The standard shape is 9 to 15 mailboxes on 3 to 5 domains with a specialist provider, which is the default for a founder-led agency running email plus their own LinkedIn profile. The multichannel shape is 27 or more mailboxes on 9 domains, which earns its cost once two or three people each work their own queue.

Ownership is the line to check before you buy, and one vendor states it plainly. Instantly's pre-warmed account documentation says the company retains domain ownership and administrator access for those purchases and is unable to transfer them. That is a fair trade for sending on day one, and it means a year of sending history, DNS and reputation sits in an account you do not control. Ourix solves B2B acquisition for founder-led B2B businesses, agencies first, by installing a done-for-you outbound system that books sales calls with their ideal clients every week without the founder prospecting. It runs on the client's own domains and accounts, with verified lists of the buyers they want, outreach on email and LinkedIn, every reply in one inbox and a pipeline view of what got booked. Campaigns go live in about 30 days, everything is the client's to take over at any time, and there is no lock-in. Ourix doesn't publish prices: the build is a one-time fee scoped to your channels and your market, and you get the exact number on the call.

When is adding more inboxes the wrong fix?

Whenever the problem is the reply rate rather than the volume, which is most of the time. More inboxes multiply whatever your current message does, so they are an amplifier and never a repair.

Run three checks before you buy anything. Bounce rate under 2 percent, which is Instantly's benchmark threshold and the signal that your verification is working. Positive replies above about 15 percent of all replies, because Agentic Demand's data has only around 14 percent of replies coming back positive at the average, and sitting well below that points at the offer or the buyer definition. And spam complaint rate under 0.10 percent, Google's recommended ceiling. If all three hold and you still want more meetings, the constraint is genuinely volume and inboxes are the answer.

If any of them fails, inboxes make the situation worse in a specific way: you spread a complaint-generating message across more domains, which converts one burned domain into several. That is the mechanism behind the opening scene on this page. The setup was not undersized. It was sending a message nobody wanted at a volume per mailbox that looked mechanical, and buying four more domains added capacity to the one thing that needed less of it.

There is a sequencing version of the same point. Promethean Research's 2026 survey of 119 agency owners has sales, profit margins and lead generation as the three biggest concerns of the year, with average after-tax net margin at 13 percent. At 13 percent, an infrastructure bill is noise and a quarter of wasted sending attention is not. The decision about whether a system or a person should own the sending is a separate one, covered in the build-or-hire comparison.

What breaks a setup that was sized correctly

Five failures account for nearly every collapse of a correctly sized build, and all five are process rather than arithmetic.

1. No verification before sending. Every address gets verified before it touches a mailbox, because bounces above 2 percent damage domain reputation faster than volume does.
2. Authentication half-finished. SPF, DKIM and DMARC on every sending domain, with valid forward and reverse DNS and TLS in transit, which is Google's baseline for all senders and Microsoft's requirement for anyone sending at volume.
3. Warmup switched off after launch. The engagement signal that keeps an inbox healthy comes from the warmup pool, and removing it mid-campaign is how a month-three decline starts.
4. No replacement cadence for burned domains. One domain will go bad. Keep two spares warming at all times so a failure costs you a day of capacity rather than a quarter.
5. Nobody owning the inbox. The sends are the easy half. A reply that waits four hours is usually a meeting lost, and a sequence that keeps running after someone replies on LinkedIn is worse than no sequence.

The fifth one undoes more sizing work than the other four combined.

What to do next

1. Write down the meetings you want a month. One number, and make it a target you would actually staff for rather than an ambition.
2. Run the six divisions from the figure above using 0.19 percent as the meetings-per-contact rate. You now have an inbox count and a domain count instead of a guess.
3. Buy the domains today and start warmup, because that is the three-week clock and everything else can happen alongside it. Nine .com registrations is about $126 a year.
4. Pick the provider row that matches your mailbox count. Google Workspace under ten mailboxes, a specialist per-inbox provider above it.
5. Set every mailbox to 25 a day with warmup left on, and leave it there until you have 60 days of your own bounce, complaint and reply data.
6. Spend the warmup window on the offer and the buyer definition rather than on tooling. That is the work that decides whether you need 17 mailboxes or 60.
7. At day 60, replace the benchmarks on this page with your own numbers and resize once. Two spare domains stay warming permanently after that.

If you would rather have the meeting math, the buyer definition and the sizing run with you before you spend a cent on domains, the free growth audit does that, and the output is yours either way.