The Cold Email Deliverability Tactic Most Guides Skip: Mailbox Rotation

I build and run cold email infrastructure for B2B companies, and I sent well over a million emails of my own last year. Most of what gets written about deliverability is either the same five basics repeated endlessly, or guesswork dressed up as certainty. Neither is useful on its own, so here is the honest version: the basics you actually need, and then the one habit that separates infrastructure that holds up for a year from infrastructure that quietly degrades after month three.

Deliverability is a gray zone, and that matters

Nobody outside Google, Microsoft and a handful of smaller providers knows exactly what trips a spam filter. That is not a disclaimer, it is the actual situation. Anyone who tells you they have cracked it, or that their infrastructure lands in the inbox every single time, is telling you something that is not possible to know. The filters are confidential by design, and they change without notice.

What that means in practice is that deliverability is a test and iterate exercise, not a checklist you complete once. You try an approach, you watch what happens to replies and inbox placement, and you adjust. The list below is what I would tell a client to do this month. I would not promise it holds exactly the same way in two years, because the filters will have moved by then too.

Never send cold email from your main domain

This is the first rule and the one that costs the most when it is ignored. Keep your main domain for the things that have to be trustworthy: invoices, client communication, calendar invitations. Send cold outreach from secondary domains instead, with forwarding back to your main domain so replies still reach you.

I had a client where this went wrong in a way that was hard to see coming. We booked them ten calls over two weeks, and all ten were no-shows. When we checked their calendar flow, the invitations themselves were landing in spam, because the client had been sending cold outreach from their main domain before we started working together. The domain was already burned, and it was burning their invoices and their client emails along with it. Once we moved outreach to dedicated secondary domains, the problem was gone. Your main domain does a lot of quiet work for you. Do not spend its reputation on cold outreach.

SPF, DKIM and DMARC are not optional

Before you warm up a single mailbox, every sending domain needs SPF, DKIM and DMARC configured correctly. Skip this and you are sending without authorization as far as receiving servers are concerned, and you will land in spam almost every time. This part is well understood and not really up for debate, so there is no shortcut worth taking here. I've written the full setup, record by record, in SPF, DKIM and DMARC explained for cold email if you want the exact values to copy.

Warmup now needs three to four weeks, not two

A new domain has zero reputation, so sending real volume the day after you buy it is a bad idea. Warmup tools simulate normal mailbox conversation so spam filters see activity that looks human before you start sending at volume.

What has changed is how long that takes. Two weeks used to be enough. I would not run a domain on a two week warmup anymore, because the drop in performance on a short warmup is now noticeable. Three to four weeks is where I'd set the baseline today. I cover the week by week version of this plan in how to warm up cold email domains.

Pre-warmed inboxes versus building your own

If you are an agency with a client who needs to send now, waiting four weeks is not always an option. In that case I would get pre-warmed domains and inboxes from a provider rather than starting from zero. Those often outperform infrastructure you build yourself from a fresh domain, simply because the reputation clock already started elsewhere.

The way I actually run this for clients is to use the pre-warmed inventory to start sending immediately, while warming up a custom set of domains in parallel for three to four weeks. By the time the custom infrastructure is ready, it takes over, and the pre-warmed domains have done their job of not leaving pipeline empty in the meantime.

Scale sideways, not up

Do not put all your volume through a handful of mailboxes. Spread it across more of them instead. A rough number I'd use as a starting point is twenty to thirty emails per mailbox per day, which means the way to send more is to add mailboxes, not to push each one harder. The more spread out your sending is, the less any single mailbox has to carry, and the less risk there is of one account getting suspended and taking a chunk of your pipeline with it. I lay out the actual per-provider ranges in how many cold emails per day per inbox.

Write the email like you would text a colleague

The content of the email matters as much as the infrastructure sending it. A marketing style email, logo, banner, a second signature block in the body, reads as a mass send to a spam filter and to the person receiving it. Keep it plain text, one idea, no images, no links, and no more than thirty or forty words. Something like: you just opened a Berlin office, curious how you're handling outbound for that region, worth a quick chat.

That is closer to a direct message than a marketing email, and that is the point. If you want the fuller framework for subject lines, openers and the follow-up logic around this, I go through it in how to write cold emails that get replies.

Verify the list before any of this matters

None of the above helps if the list underneath it is bad. Double or triple verify before you send, running addresses through more than one verification tool in sequence rather than trusting a single pass. A high bounce rate burns a domain faster than almost anything else on this list, because bounces are one of the clearest signals a mailbox provider has. If you want the actual sequence I use to catch what a single verification pass misses, I cover it in the email waterfall strategy.

Worth doing alongside verification: turn off open tracking and click tracking entirely. Neither helps deliverability, and the tracking pixel itself is one more thing a filter can flag.

The tactic the basics never cover: rotation

Everything above is setup. You do it once, more or less, and then it runs. The tactic that actually separates infrastructure that holds up for a year from infrastructure that quietly decays is one most deliverability guides never mention at all, because it treats mailboxes as something that gets tired rather than something you configure and forget.

Here is the system. If your volume math says you need thirty active mailboxes, provision sixty. Every two weeks, rotate: take the older half out of active sending and let it rest, while the fresh half takes over the sending load. Two weeks later, rotate again. The mailboxes that were resting come back fresher, and the ones that had been sending get their turn to recover. You can run this on a schedule without watching it daily once it is set up.

This is the difference between treating deliverability as infrastructure you install once, and treating it as a crew you manage. A crew needs rest. So does a mailbox.

The other advanced moves worth running

A few more tactics worth layering on once rotation is in place, roughly in order of how often I see people actually use them.

• Spread across five to ten providers. The same logic as mailbox rotation applied one level up. Do not run your whole fleet through one provider. Test several, keep the ones that perform, and spread your mailboxes across them so one provider's policy change does not take out your entire infrastructure at once.
• Run single note campaigns. Fewer follow-ups, sent as standalone emails rather than threaded replies, tends to perform better for deliverability than long follow-up chains. One or two touches, then move the contact to a later list rather than pushing a sixth follow-up.
• Send on weekends too. Spreading volume evenly across the whole week, instead of a nine to five pattern every weekday, looks less like an automated blast and more like normal human activity. A surprising number of people check work email on weekends, so it is not wasted volume either.
• Use spintax across quite a few subject lines. Fifteen to twenty subject line variants, spun properly so each email looks distinct, makes the whole send harder for a filter to pattern match as one campaign.
• Match sending provider to receiving provider where you can. Google to Google, Outlook to Outlook, when your sending setup allows it. Microsoft's filters are notably less forgiving of cold outreach than Google's, which is part of why some agencies skip Outlook inboxes entirely rather than fight that filter.

There is no 100%, and 95% is a good outcome

If you run every tactic above, domains, authentication, a proper warmup window, sideways scaling, human sounding copy, verified lists and active rotation, I would put my confidence at around 95% that you land in the primary inbox consistently. I would not go higher than that, and I would be skeptical of anyone who does. The remaining uncertainty is Google and Microsoft changing their filters without telling anyone, which is outside anyone's control no matter how good the infrastructure is.

That is the honest shape of this whole topic. Not a guarantee, a probability you can push up with the right habits and never all the way to certain. Treat deliverability as something you keep testing rather than something you finish.

FAQ