Cold Email Deliverability for Agencies and the Numbers That Keep You Out of Spam

Vlad Timinski· August 2, 2026· 9 min read

Every agency owner who tries cold email hits the same wall. You write what feels like a sharp message, load a list, press send, and hear nothing. The instinct is to blame the copy. The copy is usually the last thing that matters, because cold email routinely fails before a human ever reads it. It fails in the plumbing: cold email deliverability, the layer underneath the message.

I run lead generation for agencies across three channels, and I built the cold email engine that sold the bidding platform I run accounts on. Running that engine taught me one lesson I have never had to revise: deliverability is the asset. Copy can be rewritten in an afternoon. A burned sending domain takes months to recover, if it recovers at all.

One thing before the details. My published, measured evidence is Upwork, where I have analyzed 926,019 bids. Cold email has no equivalent public dataset, and I refuse to quote reply rates I have not published. So this article contains two kinds of numbers: platform rules that are publicly documented, and operating conventions careful senders converge on. I will tell you which is which every time.

Why cold email fails in the plumbing

Mailbox providers decide where your message lands before anyone opens it. Gmail and Microsoft score the reputation of your sending domain, your IP, and your behavior patterns: how fast volume ramped up, how many messages bounce, how many recipients complain, how many reply. If those signals look wrong, the message goes to spam or gets dropped silently, and your carefully written email argues its case to an empty room.

This is why "what should I write" is the wrong first question. The right first question is whether your emails arrive at all. Every channel has plumbing. LinkedIn caps connection invites at an undocumented level, commonly somewhere around 100 to 200 a week and shifting without notice, which I covered in LinkedIn outreach limits. Email's plumbing is deliverability, and most of it is inspectable if you know where to look.

Never send cold email from your main domain

Your main domain carries the business: invoices, client threads, contracts, every ordinary email you send. If cold outreach damages its reputation, all of that suffers, and you cannot start over because the domain is your identity.

So cold email runs on dedicated sending domains: lookalike variations of your brand, bought for outreach, each redirecting to your real site. If a sending domain accumulates damage, you retire it and the business is untouched. The convention most careful senders hold is a small number of inboxes per domain, usually two or three, so no single mailbox carries conspicuous volume. That is convention, not a published platform rule; treat it as a starting point.

I treat those domains as client property. Everything I build on the cold email side, the domains, the inboxes, the SmartLead campaigns themselves, transfers to the client if they ever leave. A deliverability asset you do not own is not an asset.

Cold email warmup, the weeks nobody wants to wait

A fresh domain with zero history that suddenly sends hundreds of emails looks exactly like a spammer, because that is exactly what spammers look like. Cold email warmup is the fix: weeks of gradually increasing, engagement-heavy sending before any real prospect sees a message. Warmup networks (SmartLead runs one, and most sending platforms do) exchange mail between real inboxes that open, reply, and rescue messages from spam folders, building a history of wanted email.

The convention most senders hold is two to three weeks of warmup before real volume, with warmup left running in the background afterward. That is practitioner consensus, not a measured threshold. Nobody outside the mailbox providers knows the true minimum. What I can give you is a position: I will not point a cold domain at a prospect list to save two weeks. The two weeks are cheaper than the domain.

How many cold emails per day per inbox

The range most careful senders hold is 20 to 50 cold emails a day per inbox, on top of whatever warmup traffic continues. I want to be precise about what that number is: convention, not measurement. No one has published a controlled study showing that 50 is safe and 60 is not. It is the range at which experienced operators stop seeing trouble, which is weaker evidence than a benchmark but far better than a guess presented as fact.

The important consequence is that you scale cold email horizontally, not vertically. More volume means more domains and more inboxes, warmed up in advance, never one mailbox cranked harder. When someone claims to send thousands a day from a single inbox, either they are burning it or what they run is not cold email.

SPF, DKIM and DMARC in plain English

Three DNS records authenticate your mail, and all three matter:

• SPF is a public list of servers allowed to send email for your domain. Receiving servers check whether the message came from a listed sender.
• DKIM is a cryptographic signature attached to each message, proving it really came from your domain and was not altered in transit.
• DMARC is the policy that ties them together. It tells receivers what to do with mail that fails the other two checks, and it sends you reports about who is sending as you.

Without these, you are asking Gmail to trust unauthenticated mail from an unknown domain. With them, you have merely cleared the floor. Set all three on every sending domain, then verify they pass on real delivered messages, not just that the DNS records exist. DKIM that fails silently after a platform change is a classic quiet killer.

The 2024 Google and Yahoo rules are the hard floor

In 2024, Google and Yahoo turned best practices into bulk sender requirements. The documented rules: senders over 5,000 messages a day must authenticate with SPF, DKIM and DMARC, must offer one-click unsubscribe, and must keep spam complaint rates under 0.3 percent, with under 0.1 percent as the stated safe target.

Read those complaint numbers slowly. A 0.3 percent complaint rate is three annoyed people per thousand delivered emails, and the safe target is one per thousand. That is the actual game of cold email: not slipping past filters with clever tricks, but sending relevant mail to people chosen so precisely that fewer than one in a thousand is irritated enough to click "report spam".

The 5,000 a day threshold technically exempts smaller senders. Treat the rules as your floor anyway: they describe what two of the largest mailbox providers consider legitimate mail.

Bounce rate discipline

A bounce tells the provider you are emailing addresses you never confirmed exist, which is the defining behavior of a spammer with a scraped list. Verify every list with an email verification service before it goes anywhere near a campaign, and drop catch-all and unverifiable addresses instead of gambling on them.

My operating line, which matches the convention among careful senders, is bounces under 1 percent. When a list approaches that line, I pause and clean it rather than push through. No lead list is worth a domain.

Fast reply handling protects deliverability

Here is the connection almost everyone misses: spam complaints tend to come from people who were annoyed and then ignored. Someone replies "not interested", the automation keeps following up anyway, and the third unwanted touch converts mild irritation into a spam report. Every one of those reports counts against the 0.1 percent target.

So reply handling speed is not just a sales courtesy, it is infrastructure. Any reply, positive or negative, must stop the sequence immediately. On my accounts, reply triage happens the same day. That habit exists because the goal metric is booked sales conversations with interested buyers, not raw replies, and interest decays by the hour. Its quiet second job is keeping complaint rates down.

Deliverability problem, copy problem, or list problem

When a campaign underperforms, the fix depends on which layer is broken, and the symptoms differ:

Diagnose in that order. A deliverability failure makes copy testing meaningless, because variants cannot compete for readers who never see them. Only when placement checks pass and bounces are clean does the message deserve the blame. Copy failure is its own discipline, and the patterns behind why Upwork proposals get ignored show up almost unchanged in email: generic openers, self-centered pitches, no evidence of research.

The monthly cold email deliverability health check

Deliverability is not set and forget. It drifts, and the drift is invisible until replies stop. Once a month, for every account, the check should show:

• Every sending domain warm, with warmup traffic still running in the background
• Bounce rate under 1 percent for the trailing month, the convention line described above
• SPF, DKIM and DMARC passing on real delivered messages, verified from actual headers
• Spam placement spot checks with seed addresses at Gmail and Outlook
• Complaint signals in Google Postmaster Tools comfortably under the 0.1 percent safe target
• Per-inbox daily volume still inside the conventional range, no quiet creep
• Suppression list current, so past decliners are never contacted again

Anything failing gets fixed before volume resumes. That ordering is the whole position: I will cut sending volume before I risk burning a domain, every time, because volume is replaceable next week and reputation is not.

If you are weighing this channel against the alternatives, I compared Upwork, cold email and LinkedIn side by side. Upwork is the one channel where I can hand you a baseline: across 926,019 analyzed bids, development categories average roughly one reply per 30 proposals, while a well-run account converts about one in ten or better, and you can check your own numbers against that data. Cold email has no such public benchmark, which is exactly why the discipline above matters. On a channel where nobody can tell you the average, the one edge you actually control is the plumbing your mail rides on.

Vlad Timinski

Founder of Space Sales. I analyzed 926,000+ real Upwork bids to find what actually wins work, and I build automated lead generation systems for agencies. The method is free to run on the benchmarks page.