Every hosting comparison ends the same way: two monthly prices side by side, and the smaller one wins. It is a reasonable habit and an incomplete one, because a host has two prices — the one it charges every month, and the one it charges once, on the way out.
The second price is invisible at purchase. You pay it later, mostly in hours, and usually when you are already unhappy — which is exactly why it works as a retention mechanism. A provider does not need a contract to keep you if leaving is simply too much effort to schedule.
The takeaway up front: the real cost of a host is the monthly rate plus the exit cost spread over how long you actually stay. A plan that saves a small amount monthly but takes a fortnight of engineering to leave is not cheap; it is expensive with the invoice deferred. Exit cost breaks into five parts, all of them estimable in an afternoon, and most of them controllable if you make the decisions early.
Why the exit never gets priced
Three reasons, and they compound. It arrives as one lump, much later, and budgeting handles recurring costs well and one-off future ones badly. It is mostly labour, not invoice — nobody bills you for the four days an engineer spends reconstructing a mail configuration, so it never appears in the spreadsheet where hosting costs are compared. And it is unknown at signup, because at signup the environment is small. Exit cost grows with every convenience you accept afterwards: a managed queue here, a platform-specific storage call there, a panel doing something you never wrote down.
That last point is the important one. You do not choose your exit cost when you pick a host. You accumulate it afterwards.
The five components of an exit
1. Getting the data out
The one people expect, and usually the smallest.
Outbound transfer — egress — is metered by most cloud providers and included up to an allowance by most VPS and shared hosts. A full copy of your data leaving in one burst can be an unusually large month, and it is charged on the way out, when you have the least motivation to pay it. Check the rate and the allowance before you need them, not during a migration; reading the bill line by line tells you which meter this lands on.
Two details cost more than the rate itself. Archival and cold storage tiers are cheap to fill and expensive to read — that asymmetry is the business model, and a large archive can cost more to retrieve than it cost to store for a year. And snapshots are usually not portable: a provider snapshot is an internal format that restores on that provider. The thing you thought was your backup is often just a fast rollback button. Genuine portability means a file-level or dump-level copy you hold yourself, which is the practical argument for the 3-2-1 backup approach over relying on the panel's snapshots.
2. Rebuilding the environment
Almost always the biggest line, and the one nobody counts.
The website is not the workload. The workload is the website plus everything around it that someone configured once and never documented: runtime versions and extensions, cron jobs, mail routing and authentication records, TLS issuance and renewal, firewall rules, log rotation, monitoring agents, user accounts and keys, database tuning, cache configuration, the file permission fix from the incident two years ago.
On a managed platform or a control panel, much of that was done for you, so it exists in the panel's head rather than anywhere you can copy. That is not an argument against panels — they save real hours every week — but it is where the hours come back.
Estimating this line honestly is the most useful thing you can do: walk the stack, list every configured thing, put hours against each, and multiply by a rate you would actually pay. The result is usually larger than a year of the plan.
3. Proprietary surface area
Every managed service you consume is a component you cannot copy, only rewrite: a managed database, an object store with its own API, a queue, a serverless function, an image pipeline, an identity system, redirect logic living in a dashboard.
The distinction that matters is between implementations of a standard and products of their own. A managed instance of a standard relational database is portable in principle — the data exports and imports elsewhere, and the effort is a migration. A proprietary service with its own API and semantics is a rewrite, and rewrites are estimated in weeks.
A useful test at adoption time: can I name the thing I would replace this with, and roughly what moving would involve? If you can, the lock-in is priced and acceptable. If you cannot, you are taking on an unknown liability in exchange for a convenience.
4. The contract and the licences
The paperwork line, and the easiest to check.
Prepaid terms and committed-use discounts convert a discount into a switching cost — leaving early forfeits the remainder, so the discount was partly a payment for staying. Per-server licences for control panels and commercial extras often do not transfer, so the new host means buying them again. Domains carry transfer locks and post-registration waiting periods, which is why a registrar separate from your host is worth the small extra effort.
None of these are large individually. Together they set the earliest date a move is sensible — a scheduling constraint as much as a financial one.
5. Switch risk
The cost of the move going badly — a real number, even though it is probabilistic.
DNS changes propagate at the speed of your TTL, and during the crossover both environments can receive traffic: orders landing in the old database, uploads on the old disk. New IP addresses affect mail deliverability until reputation is established. Someone competent must be awake for the cutover, and a rollback plan must exist. All of that is hours, and hours at unsociable times cost more. The mechanics of shrinking this window are in migrating a website without downtime.
When lock-in is worth paying for
The mistake in the other direction is treating all lock-in as bad and building everything from primitives to stay free. That is its own bill, paid monthly in maintenance instead of once in migration.
A rule that holds up: accept lock-in that removes recurring work; refuse lock-in that only removes one-off setup. A managed database earns its keep because it takes over backups, patching, failover and the 3 a.m. call — work that recurs forever, so paying an exit cost someday to avoid it is usually a good trade. A one-click installer saves you an afternoon once and can quietly leave you with an environment nobody can reproduce. The saving is finite; the dependency is not.
The same test applies to managed hosting generally. If the managed layer genuinely absorbs operational hours, the higher monthly rate can be the cheaper total even with a harder exit. If it is mostly a nicer interface over a standard stack, you are paying for convenience with your future mobility.
Price your own exit in an hour
Do this before you sign, and once a year after.
- Inventory everything that is not a file or a database row. Services, panel settings, DNS records, certificates, cron jobs, mail configuration, firewall rules, monitoring, secrets.
- Mark each one portable, standard-but-manual, or proprietary. Portable copies over. Standard-but-manual needs hours. Proprietary needs a replacement decision.
- Estimate hours per item and multiply by a rate you would really pay.
- Measure the data. Total gigabytes to move, and the egress rate and allowance that apply.
- Check the paperwork. Term remaining, non-transferable licences, domain lock status.
- Add a cutover allowance for the crossover window and one rollback.
Add it up, divide by the number of months you realistically expect to stay, and add that to the monthly rate. Compare hosts on that figure. It reorders shortlists more often than people expect — and it is the same discipline as sizing on the resources you actually use rather than the plan headline, covered in choosing a hosting type by fit.
Keeping the exit cheap from day one
You cannot eliminate exit cost, but you can keep it boring.
- Keep the environment in text. Provisioning scripts or configuration files that live in version control, not steps performed once by hand.
- Hold your own backups, in standard formats, somewhere the host does not control. A dump you can restore anywhere beats a snapshot that restores in one place.
- Register domains away from the host. DNS is the steering wheel of a migration; keep it in your hands.
- Prefer standards where the cost is similar. Standard database engines, standard object-storage APIs, standard web servers.
- Write down what the panel does. Every time a panel configures something for you, note what it configured.
- Keep TTLs sane so a cutover is minutes rather than days.
- Rehearse a restore elsewhere, once. The only honest test of portability is restoring somewhere that is not the provider you are leaving.
FAQ
How much does it cost to migrate hosting? Almost entirely in hours, not fees. Transfer charges and a duplicate month of running two environments are usually minor next to the labour of rebuilding configuration and testing it.
What are egress fees and when do they bite? Charges for data leaving the provider's network, metered per gigabyte above an included allowance. Day to day they track your traffic. They bite hardest during a migration, when a full copy of everything leaves at once, and when pulling data back out of an archival tier.
Is vendor lock-in always bad? No. Lock-in that removes ongoing operational work is often a fair trade. Lock-in that only saved a one-time setup step is not, because the saving ends and the dependency does not.
Are provider snapshots a backup? They are a fast rollback within that provider. They usually cannot be restored elsewhere, so a portable copy in a standard format, stored off the provider, is what makes them a real backup.
How do I compare hosts if the exit cost is invisible? Turn it into a monthly figure: estimate the total, divide by how long you expect to stay, add it to the plan price. A dearer host with a two-day exit frequently beats a cheaper one with a two-week exit.
Buy the way out, not just the way in
A hosting decision is really two decisions: what you pay each month, and what it takes to change your mind. The first is printed on the page. The second you have to work out, and it decides whether you are choosing a provider or being kept by one.
Price it once, keep it small with a few habits, and revisit it annually. To compare plans on total cost rather than headline rate, see the hosting and VPS guides at Just-Server.
