Somewhere on your hosting plan is a line about isolation, and it is the item most likely to be either wasted money or the only thing that would have helped. It is also the one the industry describes worst, so it is worth setting out plainly.
What sharing a node means
On a pooled plan your site runs on a machine alongside other customers, drawing processor time from a common allowance. Most of the time this is invisible. A brochure site serving cached pages needs a few milliseconds of CPU per request, and there is plenty to go around.
It becomes visible when a neighbour gets busy. Someone else’s traffic spike, backup job or inefficient plugin becomes your slow response. Your site did not change — the machine underneath it got busier.
What it feels like
The symptom is inconsistency rather than slowness. Median response time looks acceptable, but the slowest few percent of requests are several times worse, and the pattern moves around during the day. If you have had a site that is fast every time you check it and slow every time a client does, contention is the ordinary explanation.
This is also why “it was fine when I tested it” and “it is slow” can both be true, and why a single speed-test score is weak evidence either way.
The claim nobody should be making
You will see “dedicated CPU” on a lot of hosting pages, including, until recently, ours. It is not a real thing at this level of the market. Practically all managed hosting — ours included — runs on virtual machines whose vCPUs are scheduled by a hypervisor across other tenants. Buying a whole instance does not buy you a physical core nobody else touches.
What it does buy is that no other customer is on your machine, competing for its memory, its disk and its PHP workers. That is a genuine and measurable difference. It just isn’t the one the phrase describes, so we stopped using it.
When your own instance earns its money
When requests cannot be cached and slow ones cost you something:
- Ecommerce. Cart, checkout and account pages are per-session and cannot be served from page cache. They are precisely the requests contention affects, and precisely the ones where a delay costs an order.
- Membership, LMS and community sites. Logged-in traffic bypasses page cache for the same reason.
- Anything with a deadline. Booking, ticketing, enrolment. Traffic arrives as a spike, and the spike is the entire point of the site.
When it does not
A marketing site, portfolio or blog — anything whose pages are identical for every visitor — spends nearly all its time serving cached responses that never reach PHP. Isolation improves a number you were not waiting on. A CDN, smaller images, or fewer plugins will do more for the same money.
Where the line sits here
We publish which side of it each plan is on. Starter, Standard, Plus, WP Starter and WP Pro share a node, with per-account CPU, memory and worker quotas so a neighbour cannot take more than their share. From WP Studio up — and on every Agency plan — the instance is yours alone.
If you are on a pooled plan and seeing the variance described above, that is the upgrade which addresses it. If you are not seeing it, the upgrade will not make your site faster, and we would rather not sell it to you on that basis.