Thirty sites to host and I still don't know what actually counts as a footprint

Infra side of this I can do in my sleep. The SEO side is where I don't trust my own instincts, so I'd rather ask before I build it than rebuild it in six months.

Thirty sites. Mix of aged domains and a few fresh ones, all pointing at two properties eventually. WordPress, nothing exotic, maybe 200 visits a day across the whole set so load is irrelevant. My default would be one box, thirty vhosts, thirty IPs off a /27, because that is trivially easy to manage and I can rebuild the whole thing from config in an afternoon.

Everyone in every guide says don't. Nobody says which part of "don't" is load bearing. Is the thing being detected the IP, the ASN, the nameservers, the fact that all thirty renew on the same day, or something else entirely that I'm not thinking about because I think like a sysadmin?

FWIW I care about this being maintainable at 3am more than I care about it being clever. Thirty control panels at thirty providers is not a system, it's a hostage situation.

What are you actually running?

the part that's load bearing is that you'll put all thirty behind one cloudflare account and undo the whole exercise before you've finished it

Depends what you're running, and you've almost told us but not quite. Are these thirty sites you'd be happy to show a client, or are they thirty sites whose entire purpose is a link?

If it's the first, one box is fine and this whole conversation is theatre. If it's the second then the IP is the least of it. In order of what I'd actually spend effort on:

1. Registrar and renewal batching. Thirty domains bought in one session at one registrar on one card, all renewing on the same day, is a stronger tie than an IP will ever be and it's the one nobody bothers to break up.
2. Nameservers. If all thirty point at the same two custom NS records you have rebuilt the footprint at the DNS layer for free.
3. Analytics, tag manager, ad network IDs, the same theme with the same untouched footer.
4. IP.

I sell hosting and I'm telling you the hosting is fourth. Spread across four or five providers because it costs you almost nothing to do, not because number four is where the risk is.

Second one. They exist for links, I'm not pretending otherwise.

Registrar point taken and that's exactly the kind of thing I'd have missed — the domains came in over about nine months but I did consolidate them into one account last year because managing four logins annoyed me. So I've already done the bad thing for convenience reasons.

Has anyone in this thread actually seen a network go down where the shared IP was demonstrably the cause? Genuine question. I've seen the claim a hundred times and never the evidence.

No, and you won't get one, because it isn't how it works and hasn't been since about 2014.

I ran 60 sites on two boxes from 2013 to 2019. Same host, sequential IPs, same registrar for about forty of them. Lost four sites in six years and all four went because I'd been lazy with anchors on the same money page, not because anyone drew a line between the IPs. Meanwhile I watched people spend a fortune on "class C diversity" — remember when that was the phrase everyone used? — and get flattened anyway, because they'd built thirty sites that all linked to the same two properties and nothing else. That's the footprint. The pattern of who links to whom. The hosting is where people put the effort because the hosting is the part you can buy.

Spread it across a few providers if it makes you feel better. It's cheap. Just don't kid yourself that it's the thing keeping you alive, and don't let it eat the attention that should be going into making thirty sites look like thirty unrelated sites on the page.

I disagree with the post above in emphasis rather than substance, and I think the difference matters for how you build this, so I'll lay out how I'd actually do thirty.

What you are defending against

Not a live algorithm that clusters by IP. That is the 2014 version of the threat and it's mostly gone. What you're defending against is a human, or a human-assisted review, that starts from one flagged site and follows whatever the cheapest link between properties is. Every shared attribute you leave in place is one hop on that walk. The question is not "does a shared IP get me penalised", it's "if somebody starts at site 12, how many of the other 29 do they reach before it stops being worth their time".

That reframing changes what you spend money on, because a hop is a hop whether it's an IP or a Google Analytics property ID.

The hop list, roughly in order of how cheap it is to follow

1. Identical analytics or ad IDs. Free to search, instant, complete. Do not do this.
2. WHOIS, where it isn't privacy-protected, and the historical records where it once wasn't.
3. Registrar plus creation and expiry dates. Thirty domains created within a week and expiring on the same day is a signature.
4. Nameservers. Custom NS on your own domain across all thirty is the same mistake as one IP, made at a layer people forget to check.
5. The reverse IP. Real, cheap to look up, but it's fourth or fifth on the list, not first.
6. Template and content fingerprint. Same theme, same plugin set, same about page structure, same three stock photos.
7. Outbound link graph. Thirty sites, two destinations, no other outbound links of consequence.

What I'd build for thirty

Five providers, six sites each. Not thirty providers — you were right that thirty control panels is unmanageable and an unmanaged network dies of neglect long before it dies of detection. Five is enough that no single lookup returns the set, and few enough that you can hold it in your head.

Registrars: three minimum, and stagger renewals across the calendar. This is the cheapest win available to you and you've already broken it, so unwind that first.

DNS: use each provider's own nameservers rather than one custom set. It's less tidy and that is the point.

One caveat I'll flag because I have not tested it properly: I don't know how much the CDN layer collapses all of this. If thirty sites sit behind one account at one CDN, whether that reintroduces the association in any way that matters is not something I can demonstrate either way. I keep them on separate free accounts on the assumption that it does, which may be superstition.

Summary: five hosts, three registrars, staggered renewals, no shared IDs, and spend the time you saved on making the outbound link graph look like an accident.

Five provider is good number. I do same for long time.

For this traffic you not need anything, is 200 visit. Take small vps 1gb from five different place, different country, four or five dollar each. Total is maybe 25 dollar month for thirty site. Anyway the money is not in hosting here, is in your time.

One thing nobody say yet — do not use the same ssh key everywhere and do not use same admin email on wordpress. Is not for google, is for when one site get hacked.

saving this whole thread

The shared key thing I'd already sorted, but the admin email is a good catch — I had one alias receiving all thirty because I wanted the update notices in one inbox.

Budget is about £100 a month all in, which after this thread looks like more than I need.

At £100 you're not constrained, you're choosing. Five providers at that budget gets you real boxes rather than the bottom tier, which mostly matters because the bottom tier suspends you for a bad neighbour's spam and then you're migrating six sites on a Sunday.

seven hop list and the first one is still "don't put the same analytics id on all of them" because that's the one people actually do

No argument with the list, only with where the effort goes. You've written seven items and six of them are about somebody manually walking the network, which happens to approximately nobody running thirty sites in a boring niche.

Build it that way if you enjoy building it that way. I'd still bet the thing that kills it in three years is anchor text, and none of the seven items help with that.

Both of you are describing failure modes you've never watched happen. That's the honest state of this topic.

Built it. Five providers, six sites each, three registrars, renewals spread over four months, per-provider nameservers, no shared IDs, separate admin emails, separate keys. Came in at £71 a month.

The honest answer to my own question after two weeks of doing it: the sysadmin instinct — one box, one config, rebuild in an afternoon — is the thing you're giving up, and it's a real cost that nobody in these threads prices in. My rebuild story is now five rebuild stories and I've written a runbook I'll never test.

I've gone with hexforge's layout and old_growth's priorities, which I don't think is a contradiction. Ask me in a year.