Short answer: there are three models. Resell hosting (you buy wholesale, bill the client, keep the margin, and own the support), refer it (the client buys direct and you take a commission), or pass it through and charge for your own work instead. Reselling pays the most per site and costs the most to run. Referring pays less and costs almost nothing. Which one wins depends far less on the margin than on whether you want to be a hosting support desk at 2am.
Most of what is written about this comes from companies selling reseller plans, so it tends to arrive at one answer. This is the comparison with all three options in it, including the one where you make nothing on hosting at all and are better off for it.
The three models at a glance#
| Reselling | Referring | Pass-through | |
|---|---|---|---|
| Who bills the client | You | The host | The host |
| Who answers “my site is down” | You | The host | The host |
| Typical revenue per site | $5-$30/month margin | One-off bounty, or 5-20% recurring | Nothing directly |
| Your setup cost | High: billing, support, terms | None | None |
| Client keeps the site if you part ways | Messy: you own the account | Clean: they own it | Clean: they own it |
| Cashflow risk | Yours: you pay the host regardless | None | None |
| Scales to 50 sites | Only with real operations | Yes | Yes |
| Best for | Firms already running a helpdesk | Almost everyone else | Agencies selling expertise, not infrastructure |
Model 1: Reselling#
You buy hosting wholesale, usually as a reseller plan or a block of resources, split it among clients, and bill each client at your own price. The client may never learn who the underlying host is.
The appeal is obvious. You buy at $8 a site and sell at $25, and across 40 sites that is $680 a month of recurring revenue for infrastructure you did not build. It is also the model with the longest history, which is why every control panel vendor supports it.
The costs are less obvious, and they are not mostly financial:
You become the support desk. When the client’s site goes down, they call you, because you are who they pay. That is true at 9am and it is true on Christmas Eve. You are now in the business of diagnosing whether a problem is a plugin, a theme, PHP memory, or the host, and you are doing it as the middle layer without direct access to the host’s monitoring.
You carry the billing. Failed cards, dunning, refunds, sales tax in whichever states apply to you, and chasing a client who has not paid for two months while the underlying host still bills you on time. Every one of those is a small job that arrives without warning.
You carry the float. You pay the host up front for capacity, and you collect from clients afterwards. At ten sites that is a rounding error. At sixty it is real working capital, and it is money you cannot spend on the business.
The exit is awkward. The account is yours, so when a client leaves, moving them out is a migration that somebody has to perform and somebody has to pay for. Clients who feel locked in tend to notice, and it sours otherwise good relationships at exactly the wrong moment.
Reselling is the right answer for a specific shape of firm: one that already runs a helpdesk with defined hours, already has billing operations, and is selling a managed service where hosting is genuinely one component rather than the product. If you are an MSP with a ticketing system and someone whose job is billing, the margin is real and the overhead is already paid for. If you are three people who build websites, you are about to acquire a second business you did not plan for.
Model 2: Referring#
The client buys hosting directly. You send them, and the host pays you a commission.
You give up the markup. In exchange you give up the entire operational load: no billing, no support tier, no float, no lock-in, no awkward exit. The client’s contract is with the host, and if they leave you they keep their site and their account without anyone performing a migration.
The economics are worse per site and often better in total, because the model costs you nothing to run. Twenty referred sites earning a recurring share is revenue that arrives whether or not you did any work that month, and it does not consume a single hour.
The variable that decides whether referring is worth doing at all is what the commission actually pays on, and programmes differ enormously:
One-off bounty only. A flat payment when the customer first pays. Common, simple, and it means your work is worth something once and nothing afterwards. A client who stays eight years pays you the same as one who leaves after four months.
Recurring share. A percentage of every invoice, including renewals. This is the one that compounds, and it is the difference between a referral link being pocket money and being a revenue line.
Recurring with a window. A percentage, but only for 12 or 24 months. Better than a bounty, worse than it looks, and the cut-off is usually in the terms rather than on the marketing page.
Attribution rules. How the host decides a customer is yours. Most use a cookie with an expiry, which means a client who clicks your link and signs up three months later is not yours. Some allow a customer to be claimed after the fact.
The last one matters more than people expect. If you have twenty existing clients you are planning to move, whether you can claim accounts retroactively, and whether a retroactive claim earns on past invoices or only future ones, is worth more than a percentage point of commission.
Model 3: Pass-through, and charge for your work instead#
The client pays the host directly, you make nothing on the hosting, and you charge properly for what you actually do.
This sounds like leaving money on the table and frequently is not. Hosting margin on a small site is $10-$20 a month. One extra billable hour is worth more than a year of it. Agencies that stop trying to earn on infrastructure and put that energy into a maintenance plan generally end up with more revenue and dramatically less operational surface.
It also has the cleanest position in a conversation. When you recommend a host and make nothing from the recommendation, the recommendation is worth more, and you can say so.
The strongest version of this model is pass-through plus referral: the client owns their account and pays the host, you take a commission for the introduction, and your invoice covers your expertise. You get the clean exit, the honest recommendation, no support burden, and a small recurring line underneath it.
There is a practical detail that decides whether clients accept it. Put hosting on the proposal as a third-party cost the client pays directly, named and priced, in the same place you would list a premium plugin licence or a stock photo subscription. Clients are used to that. What they object to is a vague “hosting and infrastructure” line at a number they cannot check, which is exactly what reselling produces and exactly the line that gets questioned at renewal. Naming the host and the real price removes the argument and makes your own fee look like what it is: the part that requires a person.
The numbers, on twenty sites#
Assume twenty client sites on a plan the client would pay about $120 a year for.
Reselling at a $15/month margin: $3,600 a year. Against that, put the support hours. Twenty sites generating even one hosting-related question each per quarter is eighty interruptions a year, and an interruption during focused work costs more than the twenty minutes it takes. If you value your time at $125/hour and each one really costs you 45 minutes end to end, that is $7,500 of time against $3,600 of revenue. The margin only works if the support is genuinely near zero or somebody cheaper than you is absorbing it.
Referring with a $30 bounty and a 15% lifetime share: $600 in bounties in year one plus $360 a year in recurring, so $960 in year one and $360 every year after, forever, for zero hours. Over five years that is $2,400 and no support load.
Pass-through with a maintenance plan at $150/month across even eight of those twenty clients: $14,400 a year, for work you were going to do anyway and can price properly because it is not muddled together with a hosting line.
The ordering is the point. Hosting margin is the smallest number in every realistic scenario, and it is the one that carries all the operational risk. Chasing it is usually the worst-paid decision on the list.
What actually decides it#
Ignore the percentages for a moment and answer four questions honestly.
Do you want to answer the phone? If the answer is no, reselling is already out. Everything else about the model is negotiable; this is not.
What happens when a site is compromised? If you resell, the incident is yours to manage, including the client conversation. Being the one who has to explain a compromise is a different job from being the one who fixes it, and both land on you. Choosing a host with serious protection at the server level rather than relying on plugins changes how often that call happens, whichever model you pick.
What happens if the client leaves? Under reselling, their site lives in your account, and separating cleanly costs somebody time and goodwill. Under the other two, they walk away with everything and you keep the relationship intact. Agencies underrate this until the first time it happens.
Can you carry the float? If you would notice paying three months of hosting for a client who has stopped paying you, do not resell.
Questions to ask before joining any programme#
Whichever direction you go, the terms matter more than the headline rate:
- Is the commission a one-off bounty, a recurring share, or both?
- If it is recurring, is there a window, and where is that stated?
- Does it pay on renewals or only the first invoice?
- Is it calculated on the invoice total or the subtotal before tax?
- How is a customer attributed to you, and for how long?
- Can you claim an existing customer, and does a claim earn on invoices already billed?
- What is the minimum before you can withdraw, and how long is commission held first?
- If you and the client part ways, whose account is it?
The last one is the one nobody asks and everyone should.
Where Hostney sits#
We do not sell a reseller tier, and the honest reason is that the model above only works well for firms already built to run a helpdesk. What we run instead is a partner programme: a $30 bounty when a referred customer pays their first invoice, then 15% of every invoice after that, renewals included, with no cut-off date. Commission is held for 90 days before it becomes withdrawable, and after that it is paid through Stripe on demand.
Two details worth knowing because they are the ones that differ between programmes. Attribution happens at signup and is permanent, so there is no cookie window to lose a slow-deciding client to. And an existing customer can be claimed after the fact with a code from their settings, though a claim only earns on invoices raised after it, which is why sending the link before a client signs up is worth real money.
The client owns their account and their billing throughout. If you want to work inside it, delegate access is a separate thing they grant you, and it does not depend on the referral. That combination is deliberate: you get the recurring line without becoming the support desk, and the client keeps a site they can walk away with. There is a free trial if you want to put a real project on it before you recommend it to anyone, which is the order we would suggest.
Summary#
Reselling hosting pays the most per site and is a business, not a revenue stream. It suits firms with a helpdesk and billing operations already in place, and punishes everyone else with support load, float, and lock-in that surfaces at the worst possible moment.
Referring pays less, costs nothing to run, and compounds if the programme pays on renewals rather than only on signup. Check the attribution rules and the recurring window before the headline percentage.
Pass-through, where the client pays the host and you charge properly for your own work, is the model most small agencies should default to, ideally with a referral commission underneath it. The hosting margin was never the money. It was always the maintenance plan and the expertise, and infrastructure revenue is a distraction that costs more hours than it returns.
Whichever you choose, the decision that matters more than any of this is which host you put clients on and how much of the operational work it absorbs. A host that handles updates, backups, and security properly makes every one of these three models cheaper to run. A host that does not will eat the margin on all three.