Somewhere in your agency there is a person who did not join to be a systems administrator, but has become one anyway. They know which client is still on PHP 7.4 and why. They keep half an eye on disk usage. They are the one who gets the Saturday message when a contact form stops sending. Their job title says designer, developer, or director — but a chunk of their week belongs to server care, and nobody is invoicing for it.
This is the quiet tax on agencies that host client sites. It rarely appears on a P&L, because it is paid in fragments — twenty minutes here, an evening there — by people whose time is supposed to be billed to client work. The industry’s answer to it is the word “managed”. The trouble is that “managed” is the least regulated word in hosting. This article is about what it should actually mean, and how to tell whether the version you are paying for means anything at all.
The work you are doing that you should not be
Be honest about the list. In most agencies that “just host a few client sites”, somebody is handling:
- Patching and versions — operating system updates on any VPS, PHP version migrations as old branches go end-of-life, keeping ahead of the WordPress core update that breaks a legacy theme.
- Security response — the compromised plugin, the defacement, the mysterious outbound spam that gets the server’s IP blocklisted and takes every client’s email down with it.
- Email plumbing — SPF records that stopped passing when the client added a new marketing tool, DKIM keys that were never rotated, DMARC reports nobody reads.
- Backups — or more precisely, the assumption that backups are happening, tested by nobody, until the day a restore is needed urgently.
- Monitoring — which in practice means “the client phones when it is down”.
None of this is agency work. All of it is genuinely skilled work — which is exactly why doing it half-attention, between client projects, is the worst of both worlds: it consumes your most expensive people and still is not done to the standard a specialist would hold.
What “managed” should actually include
When a host says managed, ask them to be specific. Here is the checklist worth holding against the marketing page.
1. The platform is patched by the platform’s engineers
On managed shared and reseller hosting, the operating system, the web server, the PHP runtimes, and the panel itself should be updated by the host, on the host’s schedule, without you being asked to do anything or even to care. You choose a PHP version per site from a menu; keeping those versions secure is not your job. If a “managed” plan still leaves OS patching to you, it is an unmanaged plan with better fonts.
2. Isolation that contains problems
Management is not only maintenance — it is architecture that limits blast radius. Each client account should be isolated from the others, so that one site’s traffic spike, infinite loop, or compromised plugin cannot degrade its neighbours. On FXRM’s Enhance-based platform, per-client resource isolation is standard: the misbehaving site is the misbehaving site’s problem, not your whole book’s. Without isolation, every incident is potentially a nineteen-client incident, and “managed” mostly means “we will apologise centrally”.
3. Email that is configured correctly by default
The single most common cause of “hosting” complaints is not hosting at all — it is email deliverability. A managed platform should configure SPF, DKIM, and DMARC correctly on every domain from day one, run real spam and antivirus filtering, and keep the small details right: reverse DNS on the sending IPs, proper TLS on the mail paths, sender reputation actively looked after. FXRM runs its own Zimbra mail infrastructure with authentication configured on every domain as standard, not as a chargeable add-on or a knowledge-base article you are left to follow. If your host treats DNS and mail records as your responsibility, you are the sysadmin again.
4. Backups with restores, not backups with promises
Everyone claims backups. The questions that separate managed from theatre: How often? Retained how long? Restorable by whom, how fast, and at what cost? A managed service means a restore is a routine operation — a file, a database, or a whole site brought back promptly — not a ticket that opens a negotiation. Ask a prospective host to describe their last real customer restore. The quality of the answer tells you everything.
5. Someone is watching who is not you
Monitoring means the host knows about a problem before your client does. Service and capacity monitoring — disk, queues, load, certificates approaching expiry — should be running continuously on the host’s side, with engineers responding to it. “We noticed and fixed it overnight” is the sentence you are paying for.
6. Escalation that reaches an engineer
The real test of managed hosting is not the feature list; it is what happens at 10pm on a Friday when a client’s site is down before a Monday launch. Is there a phone number? Does someone technically capable answer it? Are they able to actually operate the infrastructure, or are they a tier-one script whose only power is to create a ticket for the people who can? FXRM’s support is answered by the engineers who run the platform — the people with root, not a script between you and them. When you evaluate a host, evaluate that path, because you will be standing on it at the worst possible moment.
What stays yours
Honesty matters in both directions. Managed hosting does not absorb everything, and a host that claims it does is overselling. The application layer stays with you: the WordPress plugins you choose, the custom code you deploy, the content, the client relationship. A good host makes that layer easier — staged PHP upgrades, one-click staging, sensible defaults — but your agency’s craft remains your agency’s craft. The dividing line is simple: you look after what makes each client’s site theirs; the host looks after everything that makes it run. Every hour of infrastructure work you are currently doing on the wrong side of that line is margin leaking out of your business.
When a client genuinely needs more than shared hosting
There is a legitimate exception to all of this: the client whose application really does need root access, a dedicated resource envelope, or software the shared platform does not run. The trap is treating that as the moment your agency finally must become a sysadmin. It is not — it is the moment to check whether your host can scale with you rather than making you change provider.
FXRM’s answer is Proxmox-backed KVM VPS add-ons alongside the shared platform: real dedicated resources, full root when you need it, on the same account, the same support relationship, and the same engineers. The client who outgrows shared hosting moves up without moving out — and if you do take root on a VPS, you are choosing the scope of what you manage deliberately, with the platform’s engineers still behind you, instead of inheriting a whole operating system by accident because your shared host had no upgrade path.
The commercial argument
Put numbers on it. If server care consumes even five hours a month of a senior person’s time — a conservative figure for an agency hosting twenty sites on a platform that leaves patching, mail records, and incident response to them — that is £4,500–9,000 a year of unbillable senior time, plus the opportunity cost of the client work not done, plus the tail risk of the incident that takes a week. Against that, the price difference between bare-metal-cheap and properly managed hosting is small change.
And there is a growth argument hiding in the same numbers. An agency that is not doing its own sysadmin can sell hosting confidently — white-labelled, under its own brand, at proper margins — because adding the thirtieth client does not add another fraction of a sysadmin to the payroll. The host’s engineers scale; your evenings do not.
Three questions to ask your current host
- “Who patches the operating system and PHP runtimes, and when did you last do it?” — If the answer involves you, you are the sysadmin.
- “Walk me through your last out-of-hours incident: who noticed, who fixed it, how was I told?” — If the answer is silence, the monitoring is you.
- “A client’s site needs a file from twelve days ago restored. What happens next, and what does it cost?” — If the answer is unclear, the backups are decorative.
If those three answers disappoint, the problem is not that hosting is hard. It is that you are doing someone else’s job, unpaid, with your best people. Move the job to a host whose engineers do it for a living — and put the hours back into the work your clients actually hired you for.
