Most advice about cold email domains stops at the extension. Buy .com, avoid .info, keep it short. That is fine as far as it goes, and it ignores the part that actually matters once you are buying at scale: how you buy them.
Three hundred domains registered in a single order, at a single registrar, in a single afternoon, on sequential nameservers, with names built from the same two or three words, is not three hundred independent businesses. It is one fleet, and it looks like one.
What can actually be correlated
Registrant details are mostly redacted these days, so people assume the trail goes cold. It does not. The things that stay visible, and that can be lined up across domains, are:
- Registrar and registration date. Two hundred domains created at the same registrar within the same hour have an obvious relationship.
- Nameservers and hosting. A fleet pointing at one DNS provider on one IP range is one fleet.
- The names themselves. This is the one people underrate, and we come back to it below.
- Where the domain points. A redirect is a hard link between a domain and whatever it resolves to.
To be clear about what this is: a first-party account of how we run our own provisioning, not a claim to know the internals of any particular filter. We spread the buy because the alternative concentrates every one of those signals into a single event, and there is no upside to that.
The shape of a 393 domain buy
One client needed 393 .com domains. Here is what the purchase actually looked like.
| Dimension | What we did |
|---|---|
| Registrars | Spaceship (132), Dynadot (130), Cosmotown (131) |
| Orders | Three, deliberately unequal: 54, 77 and 262 |
| Purchase runs | 36 separate runs |
| Elapsed time | 11 days |
| Batch size | Between 6 and 15, never a round number |
| Times of day | 9 different slots, from 09:15 to 20:05 |
The three were Spaceship, Dynadot and Cosmotown. There is nothing magic about that particular set. What matters is that they are genuinely separate companies rather than three brands owned by the same registrar group, and that the split is close to even so no single one holds a majority of the fleet.
The detail that does the work is the irregularity. Thirty six runs of exactly ten domains, every day at nine in the morning, is still a pattern. It is just a slower one. Batch sizes were varied, times were varied, and the three orders were sized differently on purpose, so that no registrar saw a rhythm and no single day accounted for a meaningful share of the fleet.
The first two orders cleared in four days. The largest, 262 domains, took seven days on its own. If you are planning a buy like this, budget the calendar time up front. It is the constraint, not the spend.
The naming problem nobody talks about
The usual advice is do not buy branded domains, meaning do not register 300 variations of your client's company name. That is correct and it is not the real risk.
The real risk is repetition. Branded names are simply the most obvious way to repeat yourself.
We generate names from a pool of prefixes and suffixes, and the first version of this plan capped a word family at 9 uses per prefix and 11 per suffix. That felt conservative when we wrote it. Once 144 domains were registered and we could look at the fleet as a whole, it plainly was not. Too many names shared a stem. Read down the list and you could see the generator behind it.
So we stopped mid-buy and regenerated every name that had not yet been purchased, 249 of them, against a vocabulary roughly three times wider and a much stricter cap: two domains per prefix, four per suffix. The replacement pool was also built to share no prefix or suffix with anything already registered, so the new names carried no lexical link back to the half of the fleet that already existed.
The purchased domains were left untouched. Only the plan changed. If you are running a buy of this size, look at the list as a list at least once before you finish, not only as rows in a spreadsheet.
Leave the redirects off
A lot of setup guides tell you to point each sending domain at your main website, on the theory that a domain resolving to a real business looks more legitimate than one resolving to nothing.
At fleet scale that reasoning inverts. A redirect is the single cleanest way to tie hundreds of otherwise unrelated domains back to one origin. Every domain in the fleet now demonstrably belongs to the same operation, and you have published that relationship yourself.
If a domain needs to resolve to something, give it its own minimal page. If it does not need to resolve, leave it. We do not add redirects on sending domains and we would not advise it.
When this is worth doing
Honestly: above roughly 10,000 sends a day. Below that you are buying a handful of domains and none of this applies. Buy good .com names, keep them clean, move on.
The threshold matters because the effort is real. Thirty six scheduled purchase runs across eleven days is not something you do casually, and doing it badly is worse than doing it simply. If you are not at the volume where fleet correlation is a live concern, spend the effort on list quality instead.
The checklist
- Three registrars, split roughly evenly. We use Spaceship, Dynadot and Cosmotown. Not one, and not seven.
- Spread across days, not hours. Plan for a week or more on a large fleet.
- Vary the batch sizes and the times. Uneven beats regular. A predictable schedule is still a schedule.
- No branded names, and more importantly, cap word repetition. A handful of uses per word across the whole fleet, not dozens.
- Review the list as a list before you finish buying. Patterns are invisible row by row.
- No redirects on sending domains.
- Buy .com. Everything above is wasted on an extension that carries its own reputation problem.
Want this handled for you?
Our Build and Release service stands up the whole thing: domains, mailboxes, authentication, warm-up and tracking, ready to send. You skip the scheduling and start with campaigns.
Book a call