# How Many Cold Email Inboxes and Domains Do You Need?

Size inbox capacity from the messages an approved programme actually needs, including follow-ups already in flight. Then apply the real account, campaign, provider and operating constraints. There is no universal inbox-per-domain ratio that guarantees delivery. A useful plan forecasts peak demand, preserves reply and suppression handling, and buys only the identities the business can legitimately own and operate.

Canonical URL: https://www.martechdevs.com/blog/how-many-cold-email-inboxes

By [Jafrin Akter Alo](https://www.martechdevs.com/authors/jafrin-akter-alo)

Data Engineer and Co-founder at [Martechdevs](https://www.martechdevs.com) and [Growth and Analytics Partners](https://www.growthandanalytics.com)

[LinkedIn](https://linkedin.com/in/jafrinalo)

Last updated: 2026-10-25

Next step: [Instantly Pricing: Outreach, Data and Mailbox Costs](https://www.martechdevs.com/blog/instantly-pricing)

## What is the useful answer to the inbox-count question?

The minimum count is the smallest set of usable accounts that can support the approved peak workload without exceeding any applicable limit. Domain count is a separate ownership and sending-architecture decision. Neither calculation establishes permission to send or a safe deliverability threshold.

A hypothetical programme needing 66 campaign messages on a peak working day would require three accounts if each had an approved usable campaign allocation of 24 messages that day. That example says nothing about whether those messages are wanted or whether a particular mailbox service permits the programme.

| Decision | Input you need | Mistake to avoid |
| --- | --- | --- |
| Programme demand | New contacts plus scheduled follow-ups and other legitimate traffic | Counting new contacts as the entire send volume |
| Usable account capacity | Actual limits, readiness, sharing and operating reserve | Treating a technical ceiling as a delivery recommendation |
| Domain arrangement | Clear business identity, ownership and permitted architecture | Buying lookalikes or rotating identities to ignore complaints |
| Operational capacity | People who can process replies and exceptions | Creating more mailboxes than the team can monitor |

The [cold email infrastructure guide](https://www.martechdevs.com/blog/cold-email-infrastructure) covers the broader setup. Our [GTM engineering implementation work](https://www.martechdevs.com/gtm-engineering) connects demand, recipient eligibility, sending controls and CRM feedback so an inbox plan can be verified against the actual workflow.

## Which permission and provider checks come before the arithmetic?

Confirm that the intended programme is appropriate for the chosen sending service and the recipients before calculating more capacity. Adding accounts does not change those conditions.

Google Workspace's acceptable-use policy prohibits facilitating unsolicited mass email and evading filtering capabilities. Microsoft's outbound guidance recommends suitable bulk-email infrastructure for mass mailings rather than using Microsoft 365 as an unrestricted sending platform. [Google Workspace acceptable use](https://workspace.google.com/terms/use_policy/), [Microsoft outbound-mail guidance](https://learn.microsoft.com/en-us/defender-office-365/outbound-spam-protection-about)

Document the audience, source of contact data, message purpose, recipient preferences and applicable operating policy. If that review rules out the proposed programme on the selected service, change the programme or service. Do not solve a policy mismatch by purchasing more identities.

Keep necessary account notifications separate from optional marketing. The team must know which traffic can pause when delivery deteriorates and which service-message path needs separate continuity planning.

## Which sending limits must the model distinguish?

Separate message counts, recipient counts, unique recipients, per-account limits, campaign limits and tenant-wide limits. They describe different consumption units and cannot be multiplied together as though every purchased inbox creates an independent allowance.

| Constraint | Scope | What to check |
| --- | --- | --- |
| Mailbox-provider limit | User or mailbox and the applicable time window | Message and recipient units, aliases, delegated sending and account status |
| Tenant limit | Organization-wide external recipients | Actual tenant allowance and traffic from other teams |
| Sequencer account limit | One account across its assigned campaigns | Available campaign allocation after other campaigns use it |
| Sequencer campaign limit | All accounts within one campaign | Whether it is lower than the combined account capacity |
| Schedule and pacing | Sending window and spacing | Whether nominal daily capacity can fit inside the actual window |
| Operational limit | Human reply and exception handling | Whether incoming work can be processed on time |

Google Workspace documents a standard paid-user ceiling of 2,000 daily messages and different limits for trials and mail merge. Its daily limits use a rolling 24-hour window. Aliases and delegated sends count toward the user's limits; an alias is not an independent new sending account. [Google Workspace sending limits](https://knowledge.workspace.google.com/admin/gmail/gmail-sending-limits-in-google-workspace)

Exchange Online separately documents a tenant external-recipient rate limit in a sliding 24-hour window, with the tenant's allowance available in its outbound-recipient report. External members of distribution groups count after expansion. Therefore, inbox count multiplied by a mailbox recipient ceiling can overstate the organization's usable capacity. [Exchange Online tenant and mailbox limits](https://learn.microsoft.com/en-us/office365/servicedescriptions/exchange-online-service-description/exchange-online-limits)

These ceilings are not recommendations to run campaigns at that volume. Keep the approved operating allocation below the applicable constraints and change it only from actual evidence, as described in the [email warmup readiness guide](https://www.martechdevs.com/blog/email-warmup).

## What do outreach-platform recommendations actually mean?

A platform's suggested setting concerns that product and account type. It does not override the mailbox provider or prove that a particular recipient cohort will receive the mail in its inbox.

Instantly's account-limits page distinguishes a per-account campaign limit, shared across all campaigns, from a campaign-wide limit across its accounts. Warmup mail is excluded from those two campaign counters. The page recommends 30 campaign messages per day for other mailboxes and caps AirMail at 20; a slow-ramp setting can initially permit less. [Instantly account and campaign limits](https://help.instantly.ai/en/articles/6248612-account-and-campaign-limits)

That creates three practical checks. Confirm the account type, inspect the live configured limit, and count other traffic under the mailbox provider's own rules. A warmup message being excluded from the sequencer's campaign counter does not make it irrelevant to every provider constraint.

Instantly's broader strategy article recommends three to five accounts per domain and describes 30–50 messages per account. Those are vendor recommendations, not a universal domain rule, and the more specific account-type limits still matter. Do not select a higher number from a general strategy page when the actual product configuration has a lower cap. [Instantly strategy guidance](https://help.instantly.ai/en/articles/5975326-instantly-cold-email-strategy)

A short sending window can be the tighter constraint. Instantly's troubleshooting guidance explains how schedule, timezone and gaps affect available sends. Raising the daily limit cannot create more minutes inside the configured window. [Instantly scheduling and limit troubleshooting](https://help.instantly.ai/en/articles/5975337-campaign-not-sending-complete-troubleshooting-guide)

## How do you calculate demand from new contacts and follow-ups?

Build a day-by-day forecast by cohort. Each day's new recipients create possible follow-ups on later days, subject to replies, opt-outs, eligibility and the sequence's stop rules. Those later messages remain demand even if you stop adding new contacts.

For an original hypothetical model, assume:

- Twenty working days in the planning window.
- Thirty new eligible recipients per working day, or 600 in total.
- An initial message, then possible follow-ups three and seven working days later.
- Seventy percent of each cohort expected to remain eligible for the first follow-up, and 50% for the second.
- No earlier cohorts already in flight at the beginning of the first window.

The percentages are forecast assumptions, not benchmarks or permission to ignore a current opt-out. Actual sends must use the latest eligibility and stop state.

| Sequence step | Expected messages per cohort | Cohorts whose step falls inside the first window | Messages in that window |
| --- | ---: | ---: | ---: |
| Initial message | 30 | 20 | 600 |
| First follow-up, three working days later | 21 | 17 | 357 |
| Second follow-up, seven working days later | 15 | 13 | 195 |
| Total campaign demand | 66 across a completed cohort lifecycle | Mixed cohort ages | 1,152 |

Across the full lifecycle of all 600 recipients, the expected campaign total is `600 × (1 + 0.7 + 0.5) = 1,320`. Only 1,152 fall inside the first twenty working days. The remaining 168 fall after that window: 63 first follow-ups and 105 second follow-ups.

Once the programme reaches a steady daily pattern under these assumptions, campaign demand is `30 + 21 + 15 = 66 messages per working day`. Planning from the first window's average of 57.6 would understate that ongoing demand.

If eight genuine manual reply-handling messages are also expected each day, they add 160 messages in the first window. That makes 1,312 total expected messages across the campaign and manual-reply streams. Keep the two streams separate when applying a sequencer's campaign-only limit; use the appropriate combined traffic when checking the provider's own counters.

## Can you reproduce the follow-up forecast in a small script?

Yes. The Python 3 example below calculates expected campaign demand only. It uses numbered working days, so holidays, actual calendar dates and recipient timezones must be mapped separately before using the result operationally. It does not send messages or assign permission.

```python
from fractions import Fraction

working_days = 20
new_recipients_per_day = 30
steps = (
    (0, Fraction(1)),
    (3, Fraction(7, 10)),
    (7, Fraction(1, 2)),
)

load = {
    day: Fraction(0)
    for day in range(1, working_days + max(delay for delay, _ in steps) + 1)
}
for cohort_day in range(1, working_days + 1):
    for delay, eligible_share in steps:
        load[cohort_day + delay] += new_recipients_per_day * eligible_share

inside = sum(amount for day, amount in load.items() if day <= working_days)
carryover = sum(amount for day, amount in load.items() if day > working_days)
peak = max(load.values())

assert (inside, carryover, peak) == (1152, 168, 66)
print("Campaign messages in window:", inside)
print("Campaign messages after window:", carryover)
print("Peak expected campaign messages per working day:", peak)
```

The expected shares are conditional on the sequence position, not independent events to multiply together. Here, 50% means half the original cohort reaches the second follow-up. If your data instead measures survival from the first follow-up to the second, convert that definition before substituting the number.

The calculation also assumes constant cohorts and fixed gaps. A one-day import of the entire list creates a different peak. Model the actual launch pattern, then reserve existing follow-up capacity before adding more new recipients.

## How do you turn peak demand into a minimum inbox count?

Divide peak campaign demand by the usable campaign allocation per ready account, and round up. Use an explicit operational reserve, then separately check account-sharing, campaign-wide, scheduling and provider constraints.

For equal accounts, the planning formula is:

```text
usable campaign allocation per account
  = approved campaign allocation × (1 − reserve fraction)

minimum ready accounts
  = ceiling(peak campaign messages ÷ usable campaign allocation)
```

With a hypothetical 20% reserve and the example's 66-message campaign peak:

| Assumed approved campaign allocation per account | Usable allocation after reserve | Minimum ready accounts | Combined usable campaign capacity |
| --- | ---: | ---: | ---: |
| 20/day | 16/day | 5 | 80/day |
| 25/day | 20/day | 4 | 80/day |
| 30/day | 24/day | 3 | 72/day |

These are sensitivity cases, not safe-send recommendations. Use only a row compatible with the actual product, provider terms, current readiness and observed delivery. In particular, an AirMail account capped at 20 cannot be assigned the 25 or 30 row.

For unequal accounts, add each account's available allocation instead of multiplying by an average. If three accounts have usable allocations of 24, 24 and 10, the total is 58, below the 66-message peak. Either reduce or reschedule approved demand, improve the actual constraint or add justified capacity. Do not assume the third account is ready for a larger load because the other two are.

A configured campaign limit can still be lower than the account sum. Likewise, two campaigns sharing the same accounts share the per-account allowance. The apparent capacity must be allocated once, not promised independently to every campaign.

## How many domains should hold those accounts?

Choose domains for transparent business identity, ownership and the actual architecture. There is no provider-backed universal ratio that converts three inboxes into a mandatory number of domains.

If a company's approved architecture permits at most two active sending accounts on each selected domain, three required accounts would occupy two domains. That is an inventory calculation under a chosen policy, not proof that two domains produce better delivery than one.

The questions behind that policy matter more than the quotient:

- Does every domain clearly belong to the real business and use an honest sender identity?
- Who owns registration, DNS, mailbox administration and renewal?
- Which legitimate message stream uses each identity?
- Can replies, opt-outs and complaints be reconciled across all accounts?
- Which provider and organizational limits remain shared?
- Is the team adding a useful identity or merely trying to continue traffic after a negative signal?

Google counts mail from the same primary domain together for its personal-Gmail bulk-sender classification, including subdomains. Subdomains therefore do not reset that classification. [Gmail domain-scope rules](https://support.google.com/mail/answer/14229414?hl=en)

Use company-controlled assets and real, monitored identities. Do not create invented employees or misleading lookalike domains to make the same programme appear to come from unrelated businesses. More identities also create more credentials, renewal work, failure points and reply queues.

## What should you budget besides mailbox subscriptions?

Budget the complete operating system: domains, accounts, sending software, data quality, setup, monitoring and reply handling. A low mailbox price is a small saving if every qualified conversation requires substantial research or correction.

The following monthly budget is hypothetical and uses the three-account scenario only to demonstrate cost categories:

| Cost item | Assumption | Monthly planning amount |
| --- | --- | ---: |
| Mailboxes | 3 × USD 12/month | USD 36 |
| Domains | 2 × USD 24/year, divided by 12 | USD 4 |
| Sending software | USD 60/month | USD 60 |
| Verification and other usage | USD 30/month | USD 30 |
| Recurring cash subtotal | Sum of the above | USD 130 |
| Operating capacity | 4 hours/month × USD 50/hour | USD 200 |
| Combined resource view | Cash plus valued internal time | USD 330 |

These are not current vendor quotes. Separate one-time setup and any paid implementation, and use the actual terms for the chosen tools. The [Instantly pricing guide](https://www.martechdevs.com/blog/instantly-pricing) and [platform comparison](https://www.martechdevs.com/blog/instantly-vs-smartlead) help identify the relevant commercial scope.

Also estimate the human queue. If twelve genuine responses arrive each working day and each needs an average of eight minutes to understand and resolve, that is 96 minutes of work per day. Across twenty working days it is 32 hours. That exceeds the four-hour general-operating assumption above, so reply handling needs its own capacity line rather than being hidden inside it.

Do not treat every response as a positive sales reply. Separate opt-outs, objections, out-of-office notices, wrong-person replies and qualified conversations. Each needs an appropriate action, and some must stop future sending immediately.

## What should happen when capacity or delivery changes?

Reforecast the remaining queue before changing settings. A mailbox pause, an employee's leave, a provider restriction or a successful campaign with more replies can change the available capacity in different ways.

| Change | Recalculate | Operating response |
| --- | --- | --- |
| Account becomes unavailable | Ready-account allocations and existing follow-ups | Pause or reschedule safely; avoid blindly moving the same sequence |
| More recipients reply or opt out | Remaining follow-up eligibility and human queue | Stop applicable follow-ups and assign response work |
| A provider starts deferring mail | Affected stream and outstanding retries | Investigate before adding new dispatch pressure |
| Campaign schedule shrinks | Time available for required pacing | Reduce the forecast or revise the approved schedule |
| New list is imported | Eligibility, cohort timing and peak demand | Test the incremental load before release |
| Another campaign uses the same accounts | Remaining shared account capacity | Allocate once across the whole queue |

When moving a conversation between accounts, preserve the recipient's state and the intended sender relationship. A pause should not cause another inbox to restart the sequence from its first message. Use a stable recipient and campaign key, with a single authoritative suppression and completion state.

The [Instantly implementation guide](https://www.martechdevs.com/blog/instantly-implementation-mistakes) covers reply and campaign-state failures that can make a nominally adequate mailbox plan behave badly. Use [email warmup evidence](https://www.martechdevs.com/blog/email-warmup) for readiness; elapsed time alone is not a capacity test.

## What should pass before buying more inboxes?

First prove that the programme is permitted, the demand forecast is correct and the existing setup is the actual bottleneck. Then compare additional capacity with reducing unnecessary messages or improving the workflow.

![Confirm permitted use before forecasting cohorts and follow-ups. Compare peak demand against account, campaign, tenant and operator constraints. Insufficient capacity returns to the forecast; failed reply, suppression or delivery checks return to repairing the programme, service or operating setup before approval.](https://www.martechdevs.com/assets/blog/how-many-cold-email-inboxes.svg)

Diagram connections:

- Define permitted audience sender identity and expected workload to Programme and provider conditions are satisfied?. 
- Programme and provider conditions are satisfied? to Repair the programme service or operating setup: No. 
- Repair the programme service or operating setup to Define permitted audience sender identity and expected workload. 
- Programme and provider conditions are satisfied? to Forecast new cohorts followups and separate traffic units: Yes. 
- Forecast new cohorts followups and separate traffic units to Measure ready account campaign tenant and operator capacity. 
- Measure ready account campaign tenant and operator capacity to Peak demand fits every applicable constraint?. 
- Peak demand fits every applicable constraint? to Reduce reschedule or justify additional controlled capacity: No. 
- Reduce reschedule or justify additional controlled capacity to Forecast new cohorts followups and separate traffic units. 
- Peak demand fits every applicable constraint? to Test replies suppression and delivery before expansion: Yes. 
- Test replies suppression and delivery before expansion to The operating checks pass?. 
- The operating checks pass? to Repair the programme service or operating setup: No. 
- The operating checks pass? to Approve the smallest justified setup and monitor: Yes

## Checklist

- [ ] The intended programme and sending service are compatible.
- [ ] Each sender identity and domain has a real owner and clear purpose.
- [ ] New recipients and follow-ups are forecast by sending day.
- [ ] Carryover from earlier cohorts is included.
- [ ] Campaign traffic and other messages use the correct limit counters.
- [ ] Aliases, shared campaigns and tenant constraints are not counted as independent capacity.
- [ ] Only ready accounts contribute their actual usable allocation.
- [ ] Replies and exceptions have enough human operating capacity.
- [ ] Suppression and sequence state survive pauses and account changes.
- [ ] Additional inboxes solve a measured bottleneck rather than conceal poor delivery.

For a defined infrastructure and CRM-state repair, [request a detailed GTM engineering scope](https://www.martechdevs.com/gtm-engineering#offer). If the workload or sending path is unclear, bring the campaign schedule and delivery evidence to the [$99 martech audit](https://www.martechdevs.com/pricing) before buying more accounts.

## Frequently asked questions

### Does adding an email alias create another independent sending allowance?

No. In Google Workspace, messages sent from a user's alias count toward that user's limits. Treat the alias as another address for the same underlying account rather than another ready mailbox in the capacity model. Verify the equivalent behavior for the actual provider and route. An additional From address also does not create another person to monitor replies or resolve recipient preferences. [Google Workspace limits](https://knowledge.workspace.google.com/admin/gmail/gmail-sending-limits-in-google-workspace)

### Can I stop adding new contacts and assume the campaign has stopped sending?

No. Existing cohorts may still have eligible follow-ups scheduled. Inspect the sequence queue and the platform's pause controls, then confirm which future actions remain pending. In the worked example, part of the cohort's demand falls after the first planning window. A capacity plan needs those outstanding obligations, and a recipient's reply or opt-out must update eligibility before the next dispatch.

### Should a meeting target determine how many inboxes we buy?

Only after the underlying assumptions are explicit and credible. A meeting target cannot establish reply rate, positive intent, booking rate or attendance. Work backward using measured conversion stages and realistic ranges, then check whether the required programme is appropriate and operable. If the assumptions demand implausible volume, revisit targeting and the offer rather than treating more inboxes as a guaranteed path to meetings.

### Can one shared inbox manage replies from many sending accounts?

A consolidated interface can reduce switching, but it does not remove the work or state requirements. Confirm that replies retain the correct recipient, conversation, sender and campaign context. Test opt-outs, replies from another address and account disconnections. Assign an owner and response coverage. A single screen is useful only if it reliably stops the right follow-ups and routes the right conversation.

### What should we do with unused mailboxes after a smaller pilot?

Review why they exist and which workflows, replies or recovery paths still depend on them. Stop unnecessary new activity, retain required records and transition legitimate conversations before removing access or subscriptions. Update the domain and account inventory so credentials and renewals do not become forgotten obligations. The smallest useful setup is easier to monitor than a growing collection of idle accounts with unclear ownership.

## Sources

1. [Google Workspace acceptable-use policy](https://workspace.google.com/terms/use_policy/)

2. [Microsoft outbound spam and mass-mailing guidance](https://learn.microsoft.com/en-us/defender-office-365/outbound-spam-protection-about)

3. [Google Workspace sending limits](https://knowledge.workspace.google.com/admin/gmail/gmail-sending-limits-in-google-workspace)

4. [Exchange Online mailbox and tenant limits](https://learn.microsoft.com/en-us/office365/servicedescriptions/exchange-online-service-description/exchange-online-limits)

5. [Instantly account and campaign limits](https://help.instantly.ai/en/articles/6248612-account-and-campaign-limits)

6. [Instantly strategy recommendations](https://help.instantly.ai/en/articles/5975326-instantly-cold-email-strategy)

7. [Instantly schedule and limit troubleshooting](https://help.instantly.ai/en/articles/5975337-campaign-not-sending-complete-troubleshooting-guide)

8. [Gmail sender classification and domain scope](https://support.google.com/mail/answer/14229414?hl=en)