VPS plans are described by CPU cores, memory and disk. Only one of those usually decides whether the server works, and it is rarely the one people size on.
Memory is the constraint that bites
Running out of CPU makes a server slow. Running out of memory makes it stop; the operating system terminates processes to stay alive, and what it kills is whatever was using the most, which is usually your database.
That failure is abrupt and confusing: a service disappears with no error in its own log, because it did not fail, it was killed.
So size memory for what actually runs. A database, a web server and an application each hold their own working set, and the total is what matters rather than any one of them.
CPU cores
More cores help when work happens in parallel. Several visitors at once, several jobs at once. They do not make a single slow request faster.
Most small sites are not CPU-bound at all. If yours is, find out what is consuming it before buying more, because one expensive query or one badly-behaved plugin often accounts for most of it.
Disk: size and type
NVMe and SSD storage are dramatically faster than spinning disks for the random reads a database performs. For anything database-backed, the storage type matters more than the amount.
Size for the data plus room to work. Backups, temporary files during an import, and log growth all need space, and a full disk takes down every service at once, which is the same abrupt, confusing failure as running out of memory.
Bandwidth
Usually generous enough not to matter, unless you serve video or large downloads. Check what happens when the allowance is exceeded, whether the connection is throttled or charged, because those are very different outcomes.
Managed or unmanaged
This changes the work more than any specification does.
Unmanaged means the provider supplies the machine and the network; the operating system is yours to patch, configure and recover.
Managed covers some of that, typically updates and monitoring, and costs more.
If you are not comfortable at a command line, managed is worth the difference. An unmaintained unmanaged VPS is worse than shared hosting, not better. It is a server with root access that nobody is patching.
Start smaller than you think
Scaling up is straightforward on a VPS, usually a resize and a reboot. Scaling down is often not, because disk cannot always shrink.
So over-provisioning on day one commits you to paying for it. Start at a size that fits what you actually run, watch the numbers for a few weeks, and resize when the data says to. Scaling and upgrading your VPS explains doing that.
Ahosting's VPS hosting starts at $8.79/month with dedicated resources and full root access.
Work out what you are actually short of first
If you are moving from shared hosting, measure before choosing anything.
Watch CPU, memory and concurrent requests during your busiest hour. One of them is the ceiling and the others are usually nowhere near it, and buying more of something you were not short of is the most common and most expensive mistake here.
If the honest answer is "the site is slow and I do not know why", find that out first. An uncached site on a bigger machine is still uncached.
Virtualisation type matters
Not all VPS plans give you the same thing.
KVM gives you a genuine virtual machine with its own kernel, so you can run your own operating system, load kernel modules and use containers.
Container-based plans share the host kernel. They are cheaper and lighter, and some things simply cannot be done, which you discover partway through a project rather than at purchase.
If you expect to run containers or anything unusual, check this before ordering. Understanding KVM Virtualization sets out the difference.
Ask what "dedicated" means on the plan
Providers differ on whether resources are guaranteed or shared with other virtual machines on the same host.
Dedicated CPU costs more and performs predictably. Shared or burstable CPU is cheaper and can slow down because of what someone else is doing, which is the harder problem to diagnose because nothing on your side changed.
For a production site, predictability is usually worth the difference. For a staging or development machine it is not.
Check what you can change later, and how
Most providers let you add memory and CPU. What varies is whether it needs a reboot, whether disk can be resized at all, and whether resizing down is possible once you have gone up.
That last point catches people: on many platforms you can grow a disk and never shrink it, so an over-generous choice is permanent.
Backups and snapshots are not the same
A snapshot is an image of the whole machine at a moment, taken and restored quickly, and it usually lives on the same infrastructure as the server.
A backup is a copy that exists independently. You want both: snapshots for rolling back a change you made, backups for the failure that takes the platform with it. For the arrangement, see Backing Up and Restoring Your VPS.
Check whether snapshots are included, what they cost, and how long they are kept, because those three vary enormously between providers.