Cold email deliverability is the only variable in outbound that can take a campaign from an 8% reply rate to zero without a single word of the copy changing. Same list, same offer, same sequence — but the mail is now sitting in a folder nobody opens, or being refused at the door before it ever reaches a folder. Because almost every sending tool reports a delivered rate rather than an inbox rate, most teams discover the problem months late, in the pipeline number rather than the dashboard.
This guide covers the whole stack in the order it should be fixed: what cold email deliverability actually measures, the five layers underneath it, the mailbox provider rules that now carry real enforcement, the thresholds worth holding yourself to, how to test inbox placement honestly, and a diagnosis tree for the week the numbers fall off a cliff.
What cold email deliverability actually measures
The word is used for two different things, and conflating them is the most expensive mistake in outbound. Delivery is whether a receiving server accepted the message. Placement is where that server then put it. They are not the same measurement, and only one of them predicts revenue.
| Metric | What it tells you | Where it comes from |
|---|---|---|
| Delivery rate | The message was accepted and did not bounce | Your sending tool |
| Inbox placement rate | Share of accepted mail that reached the primary inbox | Seed list testing only |
| Spam placement rate | Share routed to junk, spam or a low-priority tab | Seed list testing |
| Complaint rate | Share of recipients who hit “report spam” | Google Postmaster Tools |
| Domain reputation | The provider’s own verdict on your domain | Google Postmaster Tools |
Hold onto one fact: a 99% delivery rate is perfectly compatible with 0% inbox placement. Spam foldering is silent. Nothing bounces, no error is returned, and your dashboard stays green while the campaign quietly stops existing for its recipients. Any conversation about cold email deliverability that stops at the delivered column is measuring the wrong thing.
The five layers of cold email deliverability
Cold email deliverability problems are much easier to solve once you know which layer they live in. Every deliverability failure sits in one of five, and they have a strict dependency order.
| Layer | What it controls | Typical failure |
|---|---|---|
| 1. Identity and DNS | SPF, DKIM, DMARC, MX, PTR, MTA-STS | Authentication fails and mail is rejected outright |
| 2. Infrastructure | Sending domains, mailboxes, IPs, tracking domain | A shared tracking domain lands on a blocklist |
| 3. List quality | Verification, recency, spam traps, role accounts | A 6% bounce rate burns an otherwise healthy domain |
| 4. Content | Subject line, body, links, images, unsubscribe | Heavy HTML, shortened links, one template at scale |
| 5. Behaviour | Volume, ramp rate, complaints, replies, cadence | 300 sends a day from a three-week-old mailbox |
The order matters more than the list. Rewriting subject lines while SPF is broken is wasted work, and warming a domain that already sits on a blocklist simply warms a blocklisted domain. Work upward from layer one, and only move on when the layer below is verifiably clean.
The infrastructure spec behind good cold email deliverability
Before any copy exists, the sending estate needs a specification. Most teams improvise this, then wonder why placement is unstable across mailboxes.
- Separate sending domains. Two to four lookalike domains — getcompany.com, company-hq.com — each 301-redirecting to the main site, and none of them your corporate domain.
- Two to three mailboxes per domain. More than that concentrates risk: one complaint spike takes the whole domain with it.
- A dedicated tracking domain on each sending domain, never the shared one your tool provides by default.
- Provider limits as the outer wall, not the target. Google Workspace allows around 2,000 messages a day per user and Microsoft 365 caps recipients at 10,000 per rolling 24 hours with a 30-messages-per-minute rate limit. Cold sending should sit at roughly 1% of those ceilings.
- A real profile on every mailbox. Display name, signature, photo and a mailbox that a human occasionally reads and replies from.
Domain age matters too. A domain registered last week has no history, and providers treat the absence of history as suspicion. Buy domains at least three to four weeks before you intend to send anything from them, and warm them in that window rather than after launch.
The provider rules that now carry real enforcement
Until 2024, mailbox provider guidance on cold email deliverability was advisory. It is not any more. Three sets of rules now decide whether your mail is even considered.
- Google and Yahoo, from February 2024. Senders above 5,000 messages a day to their users need SPF, DKIM and DMARC with alignment, TLS in transit, valid forward and reverse DNS, one-click unsubscribe, and a spam complaint rate below 0.3%. Google’s own sender guidelines recommend staying under 0.10%.
- Microsoft, from 5 May 2025. Senders above 5,000 messages a day to outlook.com, hotmail.com and live.com must pass SPF and DKIM and publish DMARC at a minimum of
p=none, aligned to at least one of them. Non-compliant mail is rejected with550 5.7.515, as set out in Microsoft’s high-volume sender announcement. - One-click unsubscribe, per RFC 8058. A
List-UnsubscribeHTTPS URL plus aList-Unsubscribe-Postheader, both covered by your DKIM signature. The mailto-only version no longer satisfies Google or Yahoo.
Here is the nuance most articles skip. Those 5,000-a-day thresholds are counted per provider, so a typical outbound team running twenty mailboxes sits far below them. That does not make the rules irrelevant. They have become the filtering baseline applied to everyone, and low-volume senders who fail them get quietly filtered rather than formally rejected. Our breakdown of the Gmail bulk sender requirements walks through each rule and the bounce codes attached to it.
Cold email deliverability thresholds worth holding yourself to
Provider limits are the point at which you are punished. Operating limits should sit well inside them, because reputation falls faster than it recovers.
| Metric | Healthy | Investigate | Stop sending |
|---|---|---|---|
| Total bounce rate | Under 2% | 2–4% | Above 5% |
| Hard bounce rate | Under 0.5% | 0.5–2% | Above 2% |
| Spam complaint rate | Under 0.1% | 0.1–0.3% | Above 0.3% |
| Inbox placement rate | Above 90% | 75–90% | Below 70% |
| Reply rate | Above 3% | 1–3% | Below 0.5% |
| Sends per mailbox per day | 20–35 | 35–50 | Above 50 |
Bounce rate is the cheapest of these to control and the most damaging to ignore, because providers read it as evidence that the list was never opted in. Verification before send is therefore non-negotiable. Our guides to B2B email verification and how to reduce your email bounce rate cover the mechanics and the data decay curve behind them.
Inbox placement testing: the only honest measurement
Since spam foldering is invisible in your own reporting, the only way to see it is to own some of the receiving mailboxes. That is what a seed list is: a set of accounts across the providers your prospects use, which you send a real campaign to and then inspect by hand.
Build a seed list that matches your audience
B2B lists are dominated by Google Workspace and Microsoft 365, so a seed list stuffed with consumer Yahoo and AOL addresses tells you very little. Weight the seeds to your actual distribution, and include several genuine Microsoft 365 tenants, which filter differently from consumer outlook.com mailboxes.
Test cold email deliverability through the production path
A placement test is only valid when everything matches live sending: the same domain, the same mailbox, the same sequencing tool, the same authentication, the same tracking domain and the same links. Change any of those and you have measured a different system. Run a baseline before launch, a weekly check per domain group, and an immediate re-test after any infrastructure change.
Know what seed testing cannot tell you
Seed accounts have no engagement history with you, and providers increasingly personalise filtering per recipient. Treat a single result as directional rather than absolute, and read the trend across weeks instead. Pair it with Google Postmaster Tools, which reports complaint rates and domain reputation from real recipients rather than from seeds.
A diagnosis tree for when cold email deliverability drops
First, classify the shape of the drop in cold email deliverability. A cliff and a slide have completely different causes, and treating one like the other wastes weeks.
If it fell overnight
- Read the actual bounce codes.
550 5.7.515is Microsoft authentication.550 5.7.1is usually policy or a blocklist. The code names the layer for you. - Re-check DNS. Someone editing an SPF record, a provider rotating a DKIM key, or a new include pushing you past the ten-lookup limit will all break authentication silently.
- Check blocklists for the sending domain, the tracking domain and any redirect domain. The tracking domain is the one teams forget, and in many tools it is shared infrastructure.
- Look for a volume event. New mailboxes added to rotation, a raised daily cap, or a list import that doubled sends is the most common self-inflicted cause.
If it slid over several weeks
- Complaint rate creeping upward in Postmaster Tools, usually before anything else moves. Treat this as the leading indicator.
- List decay. Rising bounces mean the data aged out between purchase and send.
- Template fatigue. One body sent tens of thousands of times becomes a fingerprint. Vary structure, not just merge fields.
- Falling replies with flat opens. Replies drop first when placement degrades, because a promotions-tab message still records an open. Since Apple Mail Privacy Protection inflates opens anyway, cold email open rates make a poor early warning on their own.
Cold email deliverability recovery, and how long it takes
Cold email deliverability recovers far more slowly than it degrades, and the biggest mistake is resuming volume before reputation has actually moved. A workable sequence looks like this:
- Stop sending from the affected mailboxes. Not a one-day pause — a full stop until the cause is identified.
- Fix the root cause first, then re-verify the entire remaining list before a single further send.
- Resume at roughly a quarter of prior volume and ramp back over three to four weeks, watching complaint rate daily.
- Expect three to six weeks before placement normalises. Domain reputation is a trailing average, so it moves slowly by design.
- Retire the domain if reputation stays poor for more than a month after a verified fix, or if it is repeatedly blocklisted. Beyond that point you are paying rent on a bad address.
Frequently asked questions about cold email deliverability
Does the sending tool determine cold email deliverability?
Far less than vendors imply. Tools influence two things: the sending path and the shared tracking domain. Everything else — authentication, list quality, volume discipline, copy — belongs to you, and it travels with you when you switch platforms.
Should I send cold email from my primary domain?
No. Use separate sending domains that redirect to your main site. If outbound damages a reputation, it should damage one you can retire, not the domain your invoices, support tickets and marketing email all depend on.
Do open-tracking pixels hurt inbox placement?
A tracking pixel adds an image and an external domain to an otherwise plain message, which is a small negative signal. Given that privacy features have already broken open-rate accuracy, switching pixels off in cold sequences costs you a metric you could not trust and buys back a little placement.
How many mailboxes do I actually need?
Divide target daily sends by 30. Two hundred sends a day means roughly seven mailboxes spread across two or three domains. Buying more capacity than you can warm properly is how most teams create the exact problem they were trying to avoid.
Deliverability is an operating discipline, not a setup task
Most teams treat cold email deliverability as a project: configure DNS, warm the domains, launch, forget. It behaves far more like credit. It is built slowly, spent quickly, and audited continuously by parties who never explain their reasoning. Own the five layers, hold the thresholds, seed-test every week, and treat complaint rate as the number that governs all the others.
Next, go a layer deeper: email domain warmup for infrastructure, SPF, DKIM and DMARC for authentication, cold email subject lines and cold email sequence design for the content layer, or B2B data accuracy for the list underneath all of it. More in cold email and deliverability.
Sending to data you can trust is half the battle. ZenBee combines a 700M+ verified contact network with real-time buying and hiring signals and multichannel LinkedIn and email sequencing in one workflow, so outbound reaches the right people with a clean list behind it. Request a demo, or start for free.