How long should I warm a new mailbox?

4 weeks for Google Workspace and 6 weeks for Microsoft 365, warming to a >95% warmup score before you send a single cold email.

What's a healthy positive reply rate?

1% is the floor, 3-5% is healthy, and 8%+ is exceptional. Positive replies only — out-of-office, unsubscribes and auto-responses don't count.

Should I track opens and clicks?

No. Apple Mail pre-loads every pixel and corporate firewalls auto-click every link, so the numbers are fiction and the tracking hurts deliverability. Track replies — nothing else.

Best practices · the playbook

Everything we know about cold email, in six phases.

This is the end-to-end system we use to scale cold outbound to 1,000+ new prospects a day without landing in spam — planning, infrastructure, list building, copywriting, campaign setup, and the only three metrics that matter. Field-tested on real campaigns, not theorized from blog posts.

Last updated July 2026

Open the playbook Get started

Field-tested by 4,211 outbound teams sending every day

>1%

positive reply target

>95%

warmup score at launch

bounce rate target

playbook 6-phase map planning → monitoring

The 6-phase map

Field-tested

01PlanningWeek 1
02InfrastructureWks 2–5
03List buildingWks 2–5
04CopywritingWks 2–5
05Campaign setupWeek 6+
06MonitoringOngoing

Targets at launch

>1%reply

>95%warmup

0%bounce

01 / The 6-phase roadmap

One phase is a thinking week. Three run in parallel. Two run forever.

Phase 1 is planning. Phases 2–4 build at the same time while your infrastructure warms in the background — that's what turns a six-month build into a six-week one. Phase 5 launches; Phase 6 never stops. Tap any phase to open its playbook.

Week 1

Weeks 2–5

In parallel

Week 6+

Ongoing

02 / The playbook

Pick a phase. Apply the checklist.

Every phase carries the same shape — what it's for, the golden rules, the mistakes that quietly kill campaigns, and an operator checklist you run down before you move on.

phase 01 planning week 1 · the thinking week

Goal — lock the four decisions that compound through every later phase: who you're talking to, how many exist, how fast you'll reach them, and what you'll say. It's a thinking week, and skipping it is why most launches stall in Phase 4.

1A Define ICP 1B Size TAM 1C Set pace 1D Pick angle 1E Lock the Readiness Profile

Do — the golden rules

Start from customers, not aspirations. Five-plus paying customers is a pattern. Anyone naming their ICP without looking at real buyers is guessing.

Size every decision to 5,000 touches in 5 sending days. That's the statistical floor for readable signal once you're live — reverse-engineer TAM and pace from it.

Test problem-angles across segments, not subject lines. One angle on everyone is one data point. Five angles × three audiences is fifteen — from the same 5,000 touches.

Expand TAM sideways before you slow down. Too small a market? Find adjacent roles, new geographies, a reframed use case — lateral expansion beats pace compression.

Never skip the Outreach Readiness Profile. Infrastructure, lists and copy all depend on four inputs — USP, problems, case study, CTA — that only exist once the Profile is filled.

Avoid — the gotchas

Still "defining the ICP" at week three. That's a positioning problem cold email won't solve — fill the Profile and run the diagnostic instead.

Sizing for Phase 1 instead of the future feedback loop. You're planning for the moment campaigns go live in Week 6 — not for the planning week itself.

A bad ICP. Wrong audience means zero replies no matter how good the infrastructure and copy are. Fixing it after launch is expensive; getting it right in week one is free.

Phase 1 checklist Lookalike seed from real customers Reachable TAM number Daily target + runway locked 5–10 angles × 2–5 segments matrix Readiness Profile filled end-to-end

phase 02 infrastructure weeks 2–5 · warms in the background

Goal — stand up sending infrastructure designed to survive a single-point failure, and warm it to a >95% inbox-placement score before a single campaign email goes out.

2A Domain strategy 2B Account setup 2C Workspace isolation 2D Authentication 2E Warmup

Do — the golden rules

Never send from your primary domain. Buy look-alike secondaries and redirect them to your primary. A flag on an outreach domain never touches your corporate site.

Isolate every domain — one domain, one workspace. One flag on a clubbed workspace takes the whole thing down. Isolated, you replace one domain and the rest keep running.

3 accounts per domain — real humans, real faces. One-to-five (three is the sweet spot), each a real person verifiable on LinkedIn with a profile photo. The account is the fingerprint of every email.

Authenticate completely: MX + SPF + DKIM + DMARC. All four, no exceptions — they're your credentials for the internet. Missing one and you're a stranger knocking at 10 PM.

Never send from a cold account. 4 weeks warmup for Google, 6 for Microsoft, held above a 95% score. Warmup runs inside your connected tool — Instantly or Smart Lead — in the background.

Avoid — the gotchas

Cheap "authorized reseller" accounts. VPN-laundered or Education-backed accounts get nuked when Google finds out — and they always do. Ask any provider to show their reseller certificate on a call.

Role-based addresses or invented names. sales@, hello@, info@ and personas with no LinkedIn footprint all read as mass-sender.

Sending before warmup completes. Even one campaign email in week 2 of warmup can tank the domain. Respect the full 4/6-week minimum.

Silently breaking SPF via the 10-lookup limit. Adding a sender plus a warmup tool plus another app routinely pushes SPF past 10 DNS lookups — it returns permerror and a chunk of mail lands in spam while dashboards look fine. Check MXToolbox every time.

Phase 2 checklist Secondary domains purchased + redirected One workspace per domain 3 real-human accounts w/ photos MX + SPF + DKIM + DMARC verified Connected to Instantly or Smart Lead Every account >95% before Week 6

This is the phase Inboxlogy does for you. DIY infrastructure is 20–40 hours of domains, workspaces, DNS and profiles before a single account warms. Inboxlogy provisions genuine Google Workspace & Microsoft 365 accounts you own 100% — full MX/SPF/DKIM/DMARC, dedicated US and EU IPs, connected to your tool — in about a day. Warmup then runs in that tool. See pricing →

phase 03 list-building weeks 2–5 · in parallel

Goal — build a prospect list so clean and enriched that Phase 4 writes hyper-personalized copy and Phase 5 launches at 0% bounce. The list isn't "contacts" — it's mini customer briefs.

3A Define targeting 3B Source the list 3C Verify (2 gates) 3D Enrich & profile 3E Segment & load

Do — the golden rules

LinkedIn defines, other tools source. It's the cleanest professional data on the planet but can't be scraped — use it to define the ICP filter set, then pull with Apollo, ZoomInfo, a broker or Clay.

Zero bounce before a single send — verify everything. LinkedIn-verify the person, LeadMagic-verify the email. Keep safe catch-all (corporate Exchange), skip risky — blanket-killing catch-all nukes 30%+ of valid prospects.

For intent signals, scrape the primary source. Hiring → hiring portals. Local businesses → Google Maps. Running ads → Meta Ad Library. The source is always closer to the truth than a data reseller.

Make every row a mini brief. A research profile (their problems) + a personalized proposal (how you help them specifically). Phase 4 copy is only as good as this input.

One CRM, one source of truth. Sheets + HubSpot + Clay + loose CSVs is data chaos. Load everything into one CRM, tagged by segment, angle and lead magnet.

Avoid — the gotchas

Running a "clean" client list unverified. It's never clean. Every list — client, partner, yours — gets the full two-gate verification. One skipped filter has burned warmed domains for us.

Scraping LinkedIn directly. ToS violation and platform bans. Use it for definition only.

Treating Apollo/ZoomInfo data as live. It rots 5–10% a month. Re-verify against LinkedIn at time of send.

Segmenting after emails are sent. Phase 5 can't re-segment mid-campaign without breaking sequences. Segment here, in Phase 3.

Phase 3 checklist Sales Nav filter set documented Gate 1 — LinkedIn verified Gate 2 — LeadMagic checked (catch-all filtered) ~35% yield check Research + proposal on every row Single CRM loaded + tagged

phase 04 copywriting weeks 2–5 · in parallel

Goal — copy a prospect reads on their phone that maximizes positive replies while staying clear of spam filters and endpoint-security quarantines. Deliverability gets you into the inbox; copy gets you the reply.

4A Subject line 4B 4-part resonance body 4C The witty PS 4D Spintax + AI + spam-check 4E Follow-up sequence

Do — write it like a human

Lead with a problem. Pain catches attention faster than aspiration. A specific personal hook works too — never a generic "hope you're well."

Target 80 words. Ceiling 120. Send the draft to your own phone — if you have to scroll, cut. People triage email on mobile 3× more than desktop.

Everything about them, nothing about us. If a sentence could live on your About page, delete it. "You / your / their company" should dominate; "I / we / our" should be rare.

Ask one low-friction question. "Worth a look?" can be answered from a phone in five seconds. Offer a lead magnet — audit, teardown, benchmark — not a 30-minute meeting.

Spin 10–15 variations per line, AI-contextualized. Hundreds of identical emails trip ESP filters. Spintax preserves meaning while making every send unique — a deliverability control, not a nice-to-have.

Avoid — trip the filters

Zero links, images or tracking in email one. Links trigger endpoint security, images flag auto-downloaders, and even an open-tracking pixel reads as bulk-send. A link can appear in follow-up #3 — never the hook.

Don't feature-dump. Prospects buy a changed number — revenue up, cost down, time back — not a list of capabilities.

Don't ask for the calendar in email one. A meeting is the highest-friction ask in outbound — and a calendar link is a deliverability risk on a cold thread. Earn it on the second reply.

Never spin the facts. Vary the phrasing, never the claims — "{20|30|40} meetings" means two of the three are lies. And read every branch aloud before launch.

Field examples

subject & CTA

GOOD quick question, {{first_name}} Pass

GOOD Worth a look? Pass

SPIN {Hey|Hi|Hello} {{first_name}} — {quick question|one quick thing} Pass

BAD 🚀 10X Your Pipeline — GUARANTEED Flag

BAD Grab time here: calendly.com/… Flag

{{variables}} shown in sequencer syntax — swap for your tool's merge tags. Spam-check every variant before launch.

Phase 4 checklist Problem-led hook Under 80 words, mobile-checked No links / images / tracking 10–15 spintax variants per line 2–4 follow-ups, each adds value Spam-checked before send

phase 05 campaign-setup week 6+ · the launch

Goal — turn warmed infrastructure, a verified list and a 4-email sequence into live campaigns that launch at volume without burning reputation — and produce readable Week-1 signal on which angle × audience × CTA wins.

5A Load the list 5B Allocate the send 5C Human mimicry 5D Safety rails + suppression 5E Launch Week 1

Do — the golden rules

Mimic a human — no exceptions. Mon–Thu 8–10 AM prospect-local, 7–15 min random latency between sends from one account, weekend pause. A 3 AM burst is a red flag ESPs score.

One master suppression list, refreshed weekly. Customers, closed-won, competitors, in-conversation leads, opt-outs, permanent bounces — all suppressed from every campaign, before every week's launch.

Auto-terminate on reply — campaign-level AND global. The moment a prospect replies anywhere, kill their sequence everywhere. If your tool only does campaign-level, that's a tool problem — fix it before launch.

One campaign per segment — never mix. Every (audience × angle × CTA) gets its own campaign, or you can't read which combination drove the replies at the end of Week 1.

Ramp Day 1 · 50% → Day 2 · 70% → Day 3–5 · 100%. Only after warmup passes on that account. It lets real-recipient engagement build campaign reputation that warmup-pool sends can't.

Avoid — the gotchas

Launching 100% on Day 1. Even a clean account carrying full volume with no real-recipient history burns in a week. Enforce the ramp in the tool, not in your head.

A stale suppression list. Skip the weekly refresh and you'll email a customer, a competitor, or the same contact twice. The most painful Phase 5 failure — it has happened to us.

Timezone math from your own clock. Defaulting to your timezone ships LA sends into Sydney inboxes at 3 AM. Always prospect-local; flag missing-timezone rows for review.

Un-rendered spintax shipping live. One bad variant and {option1|option2} goes out to real prospects. Test-send 5 emails to yourself before every launch. Every launch.

The math that scales

10 new / account · 3 accounts / domain

Daily new (launch)	Sustained new	Accounts	Domains
300	~150	30	10
1,000	~500	100	34
5,000	~2,500	500	167

Each account sends 10 new + 10 follow-ups a day (20 total). At launch the full budget goes to new prospects; by Week 2 follow-ups stack up and consume about half. 20/day is a conservative default — clean accounts can run 30–40, but 20 keeps every behavioral signal in the green.

Phase 5 checklist One campaign per segment 5 rendered QA previews 20/day per account (10 new + 10 follow-up) Send window locked, weekend pause Global auto-terminate on reply Suppression refreshed this week Bounce circuit-breaker at 2% 50 → 70 → 100% ramp enforced

phase 06 monitoring ongoing · runs forever

Goal — watch the only three metrics that matter inside your sending tool, catch deliverability rot before it's irrecoverable, and recover domains instead of replacing them. Phase 6 isn't a setup — it's the loop you run forever.

6A The 3 metrics 6B Categorize replies 6C 10-min morning scan 6D Placement tests (on trigger) 6E Recovery protocol

Do — the golden rules

Track three metrics, ignore the rest. Warmup score (>95%), bounce rate (0%, pager at 2%), positive reply rate (1% floor · 3–5% healthy · 8%+ exceptional). Everything else is noise.

Positive reply ≠ reply. Categorize first. Out-of-office, unsubscribes, "wrong person," auto-responses — none are buying signals. Only positives count; verify the tool's categorization weekly.

Inspect daily — rot is faster than you think. A warmup drop caught Monday 10 AM is a 0-damage event. The same drop caught Friday afternoon is a 3-day infrastructure loss.

Recover before you decommission. Every domain cost 4–6 weeks of warmup. The 4-step recovery protocol saves most; only the truly broken earn a replacement.

Avoid — the gotchas

Chasing opens and clicks. Apple Mail pre-loads every pixel and corporate firewalls auto-click every link — the metric you thought you had doesn't exist, and collecting it hurts deliverability. Track replies. Nothing else.

Deleting accounts that dropped below 95%. Pause them, cool them down, re-warm slowly. A deleted account is a lost warmup investment; a paused one comes back.

Running placement tests on a calendar. When warmup is >95% and bounce is 0%, weekly tests add no information. Run them on a trigger — a warmup dip, a bounce spike, a new ESP policy — as a diagnostic.

The 10-minute morning routine

every business day

Min 0–3

Warmup-score scanSort accounts by score ascending. Anything below 95% → pause from active campaigns (keep warmup running), flag for diagnosis or recovery.

Min 3–6

Bounce-rate scan (per campaign)Check every active campaign, last 24 hours. Anything above 2% → pause and audit the list slice + suppression log before resuming.

Min 6–8

Positive-reply scanCategorize overnight replies in the unibox. Push positives to the CRM and the right rep; push negatives to suppression.

Min 8–10

The "anything weird?" checkGlance at sending volume vs plan, domain reputation on MXToolbox, auth records still valid. 90 seconds. If nothing's off, you're done.

6E the 4-step recovery protocol before you replace a domain

Audit auth records

Re-check MX, SPF, DKIM, DMARC on MXToolbox. Records valid last year can be silently invalid after a quarterly policy tightening. Fix any that fail first.

Cooldown 3–7 days

Pause all campaign sends; keep the warmup pool running at a reduced rate. ESPs need to see "slowed down after the dip," not "kept hammering."

Slow re-warmup (2 wks)

Bring warmup back up gradually, cap at 30/day during recovery. Return to campaigns only once the score holds above 95% for 3 straight days.

Decommission if needed

No recovery after 2 weeks? Retire the domain, provision a replacement, run a fresh 4/6-week warmup. Healthy accounts keep sending throughout.

Phase 6 checklist 3-metric dashboard, nothing else Every reply categorized 10-min routine owned by one person Placement tests on trigger only Recover before decommission

Where these numbers live. Warmup score, bounce rate and reply categorization are all tracked in your connected sending tool — Instantly or Smart Lead. Inboxlogy provisions and authenticates the mailboxes you monitor there; the daily scan is yours to run.

03 / The operator cheat-sheet

Write like a human. Operate like an engineer.

The defaults behind the whole playbook, on one card — the targets and windows we hold every campaign to.

The golden rule

Opens are inflated by bots and privacy proxies, click tracking plants the very links that hurt deliverability, and send volume means nothing if it lands in spam. The only metric that matters is positive reply rate. Hold it above 1% — and make every line of the system answer to that number.

Targets & windows Warm up 4 wks (Google) / 6 wks (Microsoft) Warmup score >95% Bounce rate 0% · pager at 2% Positive reply rate >1% Mon–Thu · 8–10 AM (local) Tue–Thu · 12–2 PM (local) 7–15 min between sends 10 new + 10 follow-ups / mailbox / day

Everything you need to scale

One managed stack — nothing bolted on.

Genuine Google & Microsoft mailboxes you own 100%
SPF, DKIM & DMARC — authenticated for you
Dedicated US and EU IPs for cleaner reputation
Connects to every sending tool — one click
From $2.80/mailbox · $0 setup · billed monthly

The playbook is free. The infrastructure is the hard part.

Phase 2 is 20–40 hours of DIY — or about a day with Inboxlogy. Genuine Google Workspace & Microsoft 365 mailboxes, authenticated and ready for your tool to warm.

Get started