# The Outreach Capacity Plan: Quarterly Target to Warmed Inboxes and Domains

*From one quarter's target and your closed-cohort rates, produce a dated sheet of weekly contacts, sends, warmed inboxes, domains, and per-domain complaint budgets.*

- Canonical URL: https://www.refolk.ai/guides/outreach-capacity-plan-inboxes-domains
- Pillar: Process, data, and compliance
- Format: Playbook
- Published: 2026-11-01
- Last reviewed: 2026-11-01
- Reading time: 21 min
- Keywords: how many candidates to source per hire, how many inboxes and domains for cold email, reverse funnel sourcing outreach volume, outbound capacity planning sequence steps per contact, sending domain warm up lead time planning, recruiting outreach volume per req per week

## Key takeaways

- A four-step sequence produces roughly 3.6 to 3.9 sends per contact, so follow-ups nearly quadruple the send volume a contact count implies.
- At 300 contacts a week, a 25-per-day inbox cap and 3 inboxes per domain, the plan needs 9 to 10 warmed inboxes on 4 to 5 domains including one spare.
- At about 375 sends per domain per week, Gmail's 0.3% spam ceiling is roughly 1 complaint, so a domain that gets two complaints in a week should be paused.
- Benchmark reply rates cannot be chained because their denominators differ: Gem's 16.9% is counted per person, while Belkins' 0.45% is counted per email sent.
- Gmail counts its 5,000-a-day bulk threshold across the primary domain and every subdomain, so spreading sends across inboxes or subdomains of one domain does not reset the count.
- Conservative provisioning starts about five weeks before the first cold send: a domain at least 30 days old plus 3 to 4 weeks of mailbox warm-up.

This playbook takes one quarter's hire or booked-meeting target and turns it into a dated capacity sheet. The sheet states how many people to contact each week, how many sends that creates once follow-ups are counted, how many warmed inboxes and sending domains to provision, the date provisioning must start, and the complaint and bounce budget each domain can absorb. It is written for recruiting operations, revenue operations, and anyone who will be asked how the outreach was sent. If two operators start from the same inputs, they should end up with the same sheet.

## What the outreach capacity sheet must state

A capacity sheet is a one-page document that links a quarterly outcome target to the sending infrastructure needed to reach it, with every number traceable to an input. Public tools cover only half of this each. Recruiting reverse-funnel calculators stop at "contacts needed". Cold-email inbox calculators start from a daily send number that someone made up. The sheet joins the two halves.

A finished sheet contains these fields, in this order:

- **Target and window.** Hires or held meetings, with the week each one is due.
- **Fitted rates.** Each stage rate with its sample size (n) and denominator. A denominator is the count a rate is divided by, such as people contacted or emails sent.
- **Pool split.** Expected outcomes from rediscovered contacts versus net-new contacts. Rediscovered contacts are people already in your CRM or ATS.
- **Weekly net-new contacts.** Shifted earlier by the median contact-to-outcome lag.
- **Sends.** Steady-state weekly and daily sends, with every sequence step counted.
- **Infrastructure.** Warmed inboxes, sending domains, and spares.
- **Provisioning date.** The calendar day by which domains must be registered.
- **Per-domain budgets.** Complaint and bounce counts per week, and the pause rules that apply when they are crossed.

**Outreach capacity sheet (one per quarter, per sending program)**

```
Program: ______  Quarter: ______  Owner: ______  Last re-fit: ______

1. TARGET
   Outcomes (hires / held meetings): ____
   Due by week: W1 __ W2 __ W3 __ ... W13 __

2. FITTED RATES (closed cohorts only, per person contacted)
   Contact -> interested:   ____%   n = ____   cohort dates: ____
   Interested -> outcome:   ____%   n = ____
   Rediscovered contact -> outcome: ____%   n = ____
   Median contact-to-outcome lag: ____ weeks

3. POOL
   Outcomes expected from rediscovered contacts: ____
   Outcomes needed from net-new contacts: ____
   Addressable net-new pool (de-duplicated): ____

4. VOLUME
   Weekly net-new contacts: ____   Weekly re-contacts: ____
   Sequence steps: ____   Sends per contact: ____
   Steady-state weekly sends: ____   Daily sends (÷ sending days): ____

5. INFRASTRUCTURE
   Per-inbox cold cap used: ____/day
   Inboxes required: ____   Inboxes per domain: ____
   Domains required: ____   Spare domains: ____
   Max daily sends to consumer mailboxes, per domain: ____ (must be far below 5,000)

6. DATES
   First cold send: ____   Domains registered by: ____
   Warm-up start: ____   Staggered batches: ____

7. PER-DOMAIN WEEKLY BUDGET
   Sends per domain: ____
   Complaints at 0.1% target: ____   at 0.3% ceiling: ____
   Bounces at 2%: ____   Pause rules: ____
```

*Fill every field. If a field has no source, write "assumed" and say why. A second operator should be able to rebuild every number from this page.*

## Which sending limits actually bind

Three limits bind in practice, and the published per-mailbox ceilings are not one of them. The binding limits are a practitioner cold-send cap of about 25 to 50 per inbox per day, a bulk-sender threshold of 5,000 messages a day counted per domain, and a complaint ceiling of 0.1% to 0.3%. Provider mailbox ceilings sit far above all three.

| Limit | Gmail / Workspace | Microsoft | Yahoo |
|---|---|---|---|
| Bulk threshold | Over 5,000/day to personal Gmail | 5,000+ to consumer services, per From domain | No published threshold |
| Spam / complaint | Keep under 0.10%, never 0.30% | Not covered by my sources | 0.3%, of inbox-delivered mail |
| Enforcement | Spam folder or 421 deferral | 550 5.7.515 hard rejection | Not covered by my sources |
| Mailbox ceiling | 2,000 recipients/24h (500 free or trial) | 10,000/day, 30/minute | Not covered by my sources |

Gmail's sender requirements took effect for all senders on February 1, 2024. Microsoft announced its high-volume sender rules on April 2, 2025, and began hard-rejecting non-compliant mail on May 5, 2025. Neither date changes the method. They show why you re-check provider guidance pages each quarter rather than trusting a figure copied into a spreadsheet.

#### Where the binding limits sit

1. **Provider mailbox ceiling** - 2,000 recipients a day on Workspace, 10,000 on Microsoft 365
2. **Bulk-sender threshold** - 5,000 a day to consumer mailboxes, counted per domain, not per inbox
3. **Practitioner cold cap** - About 25 to 50 cold sends per warmed inbox per day
4. **Complaint ceiling** - 0.1% target and 0.3% hard line, about one complaint a week per domain at cold volume

*The provider mailbox ceiling is the outer wall, and the limits that actually constrain a cold program sit well inside it.*

No mailbox provider publishes a cold-send cap. The 25-to-50 figure comes from practitioners and applies only after 2 to 4 weeks of warm-up. Some points are disputed and you should check them against your own tenant. Sources disagree on whether Google's external-recipient sub-cap is 2,000 or 3,000. They also disagree on Microsoft's tenant-wide external recipient limit (TERRL): one source reports that a planned 2,000-per-day mailbox external limit was cancelled, while another says a tenant-wide limit still applies. Don't size against either figure. The practitioner cap is so far below both that the dispute rarely matters.

> **Watch out:** The 5,000 count is per domain, not per inbox
>
> Gmail is reported to count bulk volume across your primary domain and all subdomains in a 24-hour window, and Microsoft counts per From domain. Spreading sends across more inboxes, or across subdomains of one domain, does not reset the count. Third-party sources also report that bulk-sender status is permanent once reached.

Recruiting programs are more exposed to the bulk threshold than B2B sales programs. The Gmail count covers only personal Gmail and googlemail addresses, and candidate outreach often goes to personal addresses. Nobody publishes what share of recruiting sends go to personal addresses, so measure it from your own send logs: count recipients by address domain for one closed month. Sources also don't establish how Gmail counts volume when several brand domains share one sender. If you run several brand domains through one setup, treat their volume as combined until you have evidence otherwise.

## How to fit stage rates without chaining benchmarks

Fit every stage rate from your own closed cohorts and express each one per person contacted. A closed cohort is a group of contacts whose sequences have finished and whose outcome window has passed, so no more replies or hires can arrive. Benchmarks are useful only as a sanity check, because each publisher divides by something different.

| Source | Rate | Denominator |
|---|---|---|
| Gem 2025 data | 16.9% reply / 9.4% interested | Per sequence (person) |
| Gem 2022 data | 21.3% / 7.9% | Per sequence |
| Instantly 2026 | 3.43% reply | Disputed: per email or per recipient |
| Belkins | 0.45% reply | Per email sent |
| Sales.co | 0.64% positive | Per contact |

A per-person rate and a per-email rate describe different events. If you multiply Instantly's reply rate (possibly per send) by Gem's per-person ATS rates, the chain looks precise and understates the volume you need. Gem's own pipeline adds a trap: it counts a sourced person as an application only after they say they are interested, so its ATS conversion rates start one stage later than most teams' funnels.

> **Rule:** One denominator per chain
>
> Every rate you multiply together must be divided by the same thing, which should be people contacted. If a rate in the chain is per email sent, convert it or drop it. A chain with mixed denominators doesn't count as a capacity plan.

When the denominators match, a chain is legitimate. Herohunt's analysis chains Gem's per-person figures and gets roughly 1,000 sequences per 2 sourced hires. It also recommends two linked funnels: sequences to interested candidates, then ATS application to hire. Build your own model the same way.

#### A same-denominator chain, Gem benchmark

| Stage | Figure | Note |
| --- | --- | --- |
| Sequences sent | 1,000 | One per person contacted |
| Replied | ~169 | 16.9% per person |
| Interested | ~94 | 9.4% per person |
| Interested in the ATS | ~80 | About 8% of sequences |
| Sourced hires | ~2 | Herohunt's chain |

*Chained on one per-person denominator, Gem's figures imply about 1,000 sequences for every 2 sourced hires.*

Three rules for the fit:

- **Close the cohort.** Include only contacts whose final sequence step has been sent and whose median lag window has passed. Open cohorts push reply rates down artificially.
- **Use recent cohorts.** Gem's reply rate fell from 22.6% to 16.9% in one year, and the interested rate fell from 12.1% to 9.4%. By the dossier's derivation, last year's rates under-provision inboxes by about 25%.
- **Treat meetings the same way.** For a meeting target, the second funnel runs from interested to held meeting. Revsculpt's estimate of 400 to 600 emails per held meeting is counted per email sent, so it can sanity-check a finished sheet but cannot feed the chain.

Two other figures help with interpretation. Direct sourcing delivers 11% of hires from 2.6% of applications, a 4x yield. Outbound offer acceptance at Gem is 73%, against 81% for inbound. If your target counts accepted offers rather than offers made, the acceptance stage belongs in your fit.

## The procedure: from target to provisioned domains

The procedure has eight steps, from a signed-off target to per-domain budgets. An analyst can complete it in about three working days, most of which goes on fitting rates. Each step has an owner, a rough duration, and a definition of done.

#### Target to provisioned infrastructure

1. **Fix the target and its date window** - RecOps or RevOps lead, about 1 hour. Done when the quarter's hires or held meetings are stated with a week-by-week due date.
2. **Fit stage rates from closed cohorts** - Analyst, 1 to 2 days. Use only contacts whose outcome is final, express every rate per person contacted, and record n and denominator beside each rate.
3. **Split the pool into rediscovered and net-new** - Sourcing lead, half a day. Done when CRM and ATS re-contacts have their own fitted rate from your own data.
4. **Back-solve weekly net-new contacts** - Analyst, about 1 hour. Divide the remaining target by the net-new rate, shift it earlier by the median lag, and check the addressable pool can supply it.
5. **Convert contacts to sends** - Analyst. Count every sequence step for net-new and re-contacted people, removing bounces and replies at each step, to get steady-state daily sends.
6. **Size inboxes and domains** - Deliverability owner. Divide daily sends by 25, place 2 to 5 inboxes per domain, add spares, and confirm every domain sits far below 5,000 a day to consumer providers.
7. **Back-date provisioning** - Count back from the first send date through domain age, DNS, mailbox creation and 3 to 4 weeks of warm-up, and write down the registration date.
8. **Set per-domain budgets and pause rules** - Fix the weekly complaint and bounce counts each domain can absorb, write the pause rules, and schedule a weekly re-fit.

Sources disagree on the order of steps 2 and 3. I fit the overall rates first, then split them. Fitting the rediscovered rate needs the same closed-cohort discipline, and doing it second lets you check that the two sub-rates recombine to the overall rate. If they don't, one of the cohorts is contaminated.

Back-solving needs the lag. If your median time from first contact to hire is several weeks, outcomes due in the quarter's last weeks need contacts sent in its first weeks or in the quarter before. Shift each week's required contacts earlier by the lag, so the sheet shows when people must be contacted, not when they are hired.

## From weekly contacts to sends, inboxes, and domains

Each contact produces roughly 3.6 to 3.9 sends on a four-step sequence, and the inbox count follows from that, not from the contact count. Replies end a sequence early, but 58% of replies arrive on step 1 and steps 2 to 4 bring in the other 42%. Most people never reply, so most people receive every step.

To calculate sends per contact, track the share of contacts still in the sequence at each step. With a 16.9% reply rate, 2% bounces at step 1, 58% of replies on step 1, and the remaining replies split evenly over steps 2 to 4 (the even split is an assumption):

- Step 1 goes to everyone: 1.000.
- Step 2 goes to the 0.882 who neither bounced nor replied.
- Step 3 goes to 0.858.
- Step 4 goes to 0.835.
- The total is about 3.57 sends per contact.

A lower reply rate means more sends per contact, because fewer people leave early. Re-contacted people from your CRM go out through the same inboxes, so add them to the weekly contact count before converting to sends. Follow-ups count against the same per-mailbox daily cap as first touches.

| Reply rate assumed | Sends per contact | Sends per 300 contacts/week | Inboxes (25/day, 5 days) | Domains (3 inboxes, +1 spare) |
|---|---|---|---|---|
| 16.9% (Gem) | ~3.57 | ~1,071 | 9 | 4 |
| 3.43% (Instantly) | ~3.87 | ~1,161 | 10 | 5 |

Every figure in this table is derived. Size for steady state. In the first week only step 1 goes out, and volume climbs as later steps start. Once the full sequence length has passed, weekly sends equal weekly contacts multiplied by sends per contact. If you size for the first week, you run out of capacity in the fourth.

**Capacity arithmetic, in the order to run it**

```
net_new_outcomes        = target_outcomes - (re_contacts x rediscovered_rate)
weekly_net_new          = net_new_outcomes_due_in_week[w + lag] / net_new_rate
weekly_contacts         = weekly_net_new + weekly_re_contacts

active[1]               = 1
active[k+1]             = active[k] - bounce[k] - reply[k] - opt_out[k]
sends_per_contact       = active[1] + active[2] + ... + active[steps]

weekly_sends            = weekly_contacts x sends_per_contact
daily_sends             = weekly_sends / sending_days
inboxes                 = ceil(daily_sends / per_inbox_cap)        # plan at 25
domains                 = ceil(inboxes / inboxes_per_domain) + spares   # 2 to 5 per domain

check: daily sends to consumer mailboxes per root domain, all senders included << 5,000
check: weekly_net_new x weeks_in_quarter <= addressable_pool - already_contacted
```

*Replace each bracketed input with your fitted value. Round inboxes and domains up, never down.*

### The supply check most plans skip

Weekly net-new contacts cannot be larger than the addressable pool can supply across the quarter. The arithmetic will happily ask for more people than exist. Here is a case from Refolk's own index.

**143 - UK profiles with Sourcer, Talent Sourcer or Technical Sourcer titles in Refolk's index**

The US figure for the same titles is 1,817, about 12.7 times larger, so a weekly volume that works for a US sourcer search can use up the whole UK pool in its first week.

A plan that calls for 300 net-new UK sourcer contacts a week fails the check before any inbox is bought. Count the pool before you commit to a weekly number, after removing everyone already in your CRM. [Refolk](/) is useful here: you describe the people you need in plain English and get back the matching profiles, so the pool size comes from real records rather than a guess.

I ran this search: `Technical sourcers in London currently at companies hiring backend engineers` - [see the full result list](https://www.refolk.ai/s/t84t5nkx4q).

*Returns matching profiles from LinkedIn, GitHub and the open web, so you can de-duplicate against your CRM and see whether the pool supports your weekly net-new number.*

## When provisioning must start

On the conservative path, provisioning starts about five weeks before the first cold send. That covers a domain at least 30 days old plus 3 to 4 weeks of mailbox warm-up, which overlap. Warm-up is the period of low, gradually rising sending that builds a mailbox's reputation before it carries cold mail. Sources put the total lead time anywhere from 2 to about 5 weeks. Plan on the long end.

#### Provisioning, back-dated from the first send

1. **Register domains** - At least 30 days before the first cold send
2. **Set DNS** - Nameserver changes take 24 to 48 hours
3. **Create mailboxes** - Ready within 24 hours
4. **Warm up** - 3 to 4 weeks, about 5 rising to 20 warm-up emails a day
5. **Ramp cold sends** - Start at 25 a day per inbox, raise only on clean budgets

*The domain's age and the mailbox warm-up run in parallel, and together they put registration about five weeks ahead of the first cold email.*

To back-date:

1. Write down the first cold send date implied by the lag-shifted contact schedule.
2. Count back 4 weeks for warm-up. That is the latest date mailboxes must exist.
3. Count back 1 day for mailbox creation and 2 days for DNS.
4. Check that the domain will be at least 30 days old on the first send date. If it won't, register it earlier.
5. If you bring mailboxes online in staggered batches, as one practitioner does over two weeks, apply steps 1 to 4 to each batch separately. Later batches can serve later weeks of the quarter.

Another source allows full volume at day 14. If you plan on that figure and your provider behaves like the 30-day source, you end up with a capacity gap of 2 to 3 weeks right when the quarter's early contacts were due.

> **Tip:** Keep the spare domain warm
>
> A spare domain is only a spare if it is already warm. Keep it sending at warm-up volume all quarter, so a paused domain can be replaced the same day instead of after a 2 to 5 week rebuild.

## Per-domain complaint and bounce budgets

Each domain gets a weekly budget written as whole counts, not percentages. At cold volumes a percentage hides how close you are to the line. At about 375 sends per domain per week (3 inboxes at 25 a day over 5 days), the budget looks like this:

- **Complaint target (0.1%):** 0.375 complaints a week, which means a normal week has none.
- **Complaint ceiling (0.3%):** about 1.1 complaints a week.
- **Bounce budget (2%):** 7.5 bounces a week.
- **Platform review line (5%):** about 19 bounces a week. MailerLite and HubSpot may suspend above 5%, and Amazon SES places accounts under review at 5%.

> At cold volumes a complaint ceiling is not a percentage. It is one person clicking a button.

One complaint in a week on a 375-send domain is about 0.27%, which is already at the ceiling. Write these pause rules into the sheet before launch:

| Signal on one domain in one week | Action |
|---|---|
| 1 complaint | Hold volume flat, review the list segment and copy |
| 2 complaints | Pause cold sends on that domain, move volume to the spare |
| Bounces above 7 (2%) | Stop the segment, re-verify before resuming |
| 421 deferrals or 550 5.7.515 | Pause and fix authentication or compliance |

Recovery takes time. If your Gmail spam rate goes above 0.3%, you lose access to Gmail's mitigation support until you keep it below 0.3% for seven consecutive days. Spreading volume across more domains contains the damage: one complaint pauses one domain out of several instead of the whole program. Watch Google Postmaster Tools for Gmail and Microsoft's Junk Mail Reporting Program (JMRP) for Microsoft complaint feedback. Yahoo's 0.3% is calculated on mail delivered to the inbox, so mail that lands in spam shrinks the denominator and inflates the rate.

## How outreach capacity plans go wrong

Most capacity plans fail in eight predictable ways, and each one has a check you can run before launch. A plan can pass every arithmetic step and still fail on one of these.

1. **Chained benchmarks.** *Looks like:* a tidy funnel built by multiplying a per-send reply rate by per-person ATS rates. It understates the volume you need. *Check:* write the denominator beside every rate in the chain. If they don't all match, rebuild the chain.

2. **Open cohorts.** *Looks like:* a reply rate that falls every time you re-run it, because recent contacts still in their sequences are counted as non-replies. *Check:* include only cohorts with every step sent and the lag window passed.

3. **Using the 46% as a contact discount.** Gem reports that 46% of sourced hires now come from rediscovered candidates, up from 26% in 2021. *Looks like:* net-new contacts cut by 46%. That is a share of hires, not of contacts. The cut assumes re-contacts convert like new contacts, and nobody has published evidence for that. *Check:* fit a separate rediscovered rate from your own cohorts.

4. **"We're under 5,000 per inbox."** *Looks like:* a sheet that tests the bulk threshold per mailbox. Gmail adds up volume across the organisational domain and all subdomains, and Microsoft counts per From domain. *Check:* add up daily sends to consumer mailboxes per root domain, across every system that sends from it.

5. **Complaint rate with a small denominator.** *Looks like:* a dashboard reading 0.00%, taken as proof of headroom. At about 375 sends a week, the next single complaint is roughly 0.27%. *Check:* track complaints as weekly counts per domain against the integer budget.

6. **"Verified" lists.** *Looks like:* a list that was verified once, months ago, or an export trusted because it came from a tool. Practitioners report 8% to 15% hard bounces on unverified exports from one popular prospecting database, which is above the 5% point where several platforms review or suspend accounts. *Check:* re-verify every segment right before launch. Apollo's guidance treats under 2% as acceptable and anything persistently above 5% as needing immediate investigation.

7. **Day-14 warm-up assumption.** *Looks like:* a provisioning date two weeks before launch. If your domains need to be 30 days old, you lose 2 to 3 weeks of capacity. *Check:* back-date on the conservative path and treat the short path as upside.

8. **Noisy title matching in the pool count.** *Looks like:* a supply check that counts every profile with a matching title. In Refolk's index, a 25-profile sample of US sourcer titles included a "President" and a fast-food employer. *Check:* sample the pool by hand before relying on headcounts, and treat title-mix percentages from small samples as low confidence.

## Who owns the sheet and how to keep it current

The sheet needs one named owner who re-fits rates every week and re-checks provider limits every quarter. In most organisations nobody has that job. Refolk's index suggests why.

**12.7 to 1 - US profiles with sourcer titles for every recruiting-operations profile, in Refolk's index**

1,817 sourcer-titled profiles against 143 with Recruiting Operations, TA Operations or Recruitment Operations titles, so the capacity model often has no dedicated owner.

| Segment | Profiles | Most common title (25-profile sample) |
|---|---|---|
| Sourcer titles, US | 1,817 | Technical Sourcer, 14 of 25 |
| Sourcer titles, UK | 143 | Talent Sourcer, 11 of 25 |
| Recruiting Ops titles, US | 143 | Recruiting Operations, 16 of 25 |

The title mix comes from 25-profile samples, so treat it as directional. The headcount ratio is the useful part: on most teams one operations person, or nobody, supports a large sourcing team. Assign the sheet to a named person, and keep the deliverability owner separate if you can.

The weekly routine:

- **Re-fit.** Add the cohorts that closed this week and recompute the per-person rates with their n.
- **Re-solve.** If the rates moved, re-run the arithmetic and compare required inboxes with provisioned inboxes.
- **Read the budgets.** Record complaints and bounces per domain as counts. Apply the pause rules.
- **Look ahead.** If next quarter's target needs more domains, check today whether its registration date has already passed.

Each quarter, re-read the Gmail sender guidelines, Yahoo's sender best practices and Microsoft's high-volume sender announcement. Check thresholds, complaint guidance and enforcement codes against what the sheet assumes. Provider rules change by announcement, and the sheet should record which version it was checked against.

#### Before you sign off the capacity sheet

- [ ] Every stage rate shows its n, its cohort dates, and a per-person denominator
- [ ] No cohort in the fit has sequences still running or an unexpired lag window
- [ ] Rediscovered contacts have their own fitted rate, and no benchmark share is used as a discount
- [ ] Weekly contacts include re-contacts, and sends count every sequence step at steady state
- [ ] Weekly net-new contacts multiplied by the quarter's weeks fit inside the de-duplicated addressable pool
- [ ] Inboxes are sized at 25 cold sends a day, rounded up, with at least one warm spare domain
- [ ] Daily sends to consumer mailboxes per root domain, across all senders, are far below 5,000
- [ ] The domain registration date assumes a 30-day-old domain and 3 to 4 weeks of warm-up
- [ ] Every list segment was re-verified right before launch
- [ ] Weekly complaint and bounce budgets are written as counts per domain, with pause rules attached
- [ ] A named owner and a weekly re-fit slot are on the calendar

## Frequently asked questions

### How many candidates do I need to source per hire?

Take the number from your own closed cohorts, not from a benchmark. As a sanity check, Herohunt chains Gem's per-person data to roughly 1,000 sequences per 2 sourced hires, because about 8% of sequences become an interested candidate in the ATS. Your role mix, seniority and message quality will move that number a lot. Fit it per role family, count only finished sequences, and state the sample size next to the rate.

### How many inboxes and domains do I need for cold email?

Divide your steady-state daily sends, follow-ups included, by a planning cap of about 25 cold sends per warmed inbox, then place 2 to 5 inboxes on each domain and add a spare. At 300 contacts a week on a four-step sequence, that comes to 9 or 10 inboxes on 4 or 5 domains, depending on whether your reply rate looks like Gem's 16.9% or Instantly's 3.43%.

### How long does it take to warm up a new sending domain?

Plan on about five weeks from registration to the first cold send. One practitioner source warms each mailbox for 3 to 4 weeks, starting at about 5 warm-up emails a day and rising to 20, on a domain at least 30 days old. Another source allows full volume at day 14. Planning on the short figure can leave a capacity gap of 2 to 3 weeks.

### Do follow-up emails count toward Gmail's 5,000-a-day bulk sender threshold?

Yes. Every message sent to personal Gmail addresses counts, and practitioners report the count runs across your primary domain and all its subdomains in a 24-hour window. Follow-ups also count against each mailbox's daily cap. A four-step sequence sends roughly 3.6 to 3.9 messages per contact, so size against total sends, not against the number of people contacted.

### Can I cut net-new sourcing because 46% of sourced hires are rediscovered candidates?

No. Gem's 46% is a share of sourced hires, not a share of contacts, and nobody publishes how well rediscovered contacts convert compared with net-new ones. If you cut net-new contacts by 46%, you are assuming both groups convert at the same rate. Fit a separate rediscovered rate from your own closed cohorts and back-solve net-new contacts from what remains of the target.

### What reply rate should I use in an outreach capacity plan?

Use your own per-person reply and interested rates, taken from closed cohorts. Public figures range from Gem's 16.9% reply rate per sequence to Belkins' 0.45% per email sent, and they use different denominators. Recruiting rates are also falling: Gem's reply rate dropped from 22.6% to 16.9% in one year, so reusing last year's rates leaves you short of inboxes.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/outreach-capacity-plan-inboxes-domains*