What an Adult Paysite Actually Costs to Run
In short. Building a paysite is a one-off number and everybody asks for it. Running one is a monthly number and almost nobody models it before launch — which is why paysites usually fail in month eight rather than month one. The costs that close them are not hosting. They are the content treadmill, the real cost of taking payment, and the churn you have to refill every single month. Below: the monthly cost structure, the numbers that decide whether the model works for you, and the point where a paysite stops being worth running.
This page is about operating economics. For structures, paywall design and launch, see how to build a profitable membership content site.
The five lines of a paysite's monthly cost
Written out, the ranking surprises most people: the technical lines are the small ones.
| Line | What drives it | How it behaves as you grow |
|---|
| Content |
Shoots, performers, editing, licensing. The largest line for almost every paysite |
Roughly fixed. It does not fall with scale, and cutting it raises churn |
| Payment |
Processing fees, refunds, dispute fees, the rolling reserve |
Rises with revenue, and the dispute portion rises faster than the rest |
| Acquisition |
Whatever it takes to replace the members who left |
Rises as the easy audience is exhausted |
| Delivery |
Storage and egress for the library |
Grows with the catalogue and with how much of it gets watched |
| Support and moderation |
Cancellations, access problems, complaints, takedown requests |
Roughly proportional to members, and cannot be dropped to zero |
Two things follow. First, a paysite is a content business with a website attached, not a website with content attached, and its cost structure says so. Second, the two lines that grow fastest — payment and acquisition — are the two least visible in a hosting invoice, which is where owners tend to look for savings.
The treadmill
A subscription is not paid for the library. It is paid for the expectation that the library will grow. The moment new material stops appearing, cancellations begin, and they begin quietly: nobody announces it, the renewal simply does not happen.
This gives the paysite model its defining constraint. Publishing cadence is a cost commitment, not a marketing choice. Before launch, the honest question is not "what can we shoot" but "what can we shoot every month for a year, at this budget, without the quality visibly dropping". A site that launches with a large library and no plan to extend it has bought itself roughly one billing cycle of goodwill.
The practical consequences are unglamorous:
- A publishing calendar exists before the site does, with the next three months already shot or contracted.
- New material is visible as new. If members cannot tell what arrived since their last visit, you are paying for content that does not reduce churn.
- Archive depth sells the first month; cadence sells every month after. Budget accordingly.
Why the real cost of payment is higher than the rate
The processing rate is the part everybody compares. It is rarely the expensive part. Four things sit on top of it.
-
Refunds cost you the sale and the fee, and in a private-purchase category the refund rate is structurally higher than in ordinary retail.
-
Disputes cost the sale, a separate fee, and a mark against the ratio card schemes monitor. Enough of them and the account itself is at risk, which is a business-ending cost rather than a line item.
-
The rolling reserve withholds a share of every settlement for a fixed period. It does not reduce revenue; it delays it, which is worse for a business that has to pay for next month's shoot out of this month's takings.
-
Failed renewals. Cards expire and get replaced constantly. A site with no retry logic and no card-update handling loses paying members it never lost the goodwill of.
⚠ That last one is the cheapest money on this page. Members lost to an expired card did not decide to leave. Recovering them costs a retry schedule and an email, against acquisition cost for a replacement member.
The two numbers that decide the model
Everything above collapses into a comparison between what a member is worth and what a member costs to acquire.
Average member lifetime (months) ≈ 1 ÷ monthly churn rate
Member value = monthly price × lifetime − payment costs − delivery and support per member
The model works while: member value > cost to acquire one member
The first line is the one worth internalising. Churn and lifetime are the same fact stated two ways: at ten per cent monthly churn the average member stays about ten months; at twenty per cent, about five. Halving churn does more for the business than halving acquisition cost, and it is usually cheaper to do, because churn is driven by cadence and by failed renewals rather than by market prices.
Two warnings about using these formulas honestly. Measure lifetime on cohorts that have actually had time to churn, not on the average of everyone including last week's joiners. And count trials as members only once they convert, or the arithmetic will flatter you at exactly the moment it matters.
Where the margin actually goes
A paysite that is busy but unprofitable almost always has one of these four:
| Symptom | Underlying cause | Where to look first |
|---|
| Revenue flat while signups grow |
Churn matches acquisition; you are refilling a bucket |
Cohort retention by month, not total member count |
| Banked revenue below reported revenue |
Refunds and disputes not counted, or the reserve not accounted for |
Your own ledger against the provider's settlement |
| Delivery cost rising faster than members |
Full-size media, autoplay previews, or no transcoding tiers |
Egress per active member, month over month |
| Support load rising per member |
Cancellation is hard to find, so it arrives as a dispute |
Ratio of cancellations to disputes |
The last row deserves a comment, because the instinct runs the other way. Making cancellation difficult looks like retention and behaves like a tax: the member leaves regardless, but does it through their bank instead of through your settings page, and the dispute costs you money and account standing. An easy cancellation is cheaper than a hard one.
When a paysite is the wrong model
Worth saying plainly, because the answer is sometimes no.
- No sustainable content cadence. Without it the model has nothing to sell after month one.
- The content is not distinguishable from what is freely available. A paywall does not create scarcity where none exists.
- No working capital for the reserve. The business is solvent and still cannot pay for the next shoot.
- The audience is a single person's following. That is a creator platform, and a different build; see migrating from OnlyFans to your own platform.
What to model before building
- Monthly content budget, committed for twelve months rather than three.
- Expected monthly churn, and the member lifetime it implies.
- Payment costs including refunds, disputes and the reserve's timing.
- Acquisition cost per converting member, by channel.
- Egress per active member, and what it becomes if usage doubles.
- Support hours per hundred members.
- The month at which cumulative revenue overtakes cumulative spend, with the reserve delay included.
How adults.dev approaches it
We build the model before the site. Content cadence, churn, payment cost and acquisition go into one sheet, and the sheet frequently changes the brief — sometimes from a subscription paysite to a pay-per-item catalogue, occasionally into the advice not to build one at all.
On the technical side the same priorities follow: a ledger of financial events that survives changing provider, retry and card-update handling so renewals are not lost to expired cards, transcoding tiers so delivery cost tracks members rather than outrunning them, and cohort retention visible on a screen instead of assembled by hand each quarter.
Model your paysite with us:
Telegram: @adultsdev
E-mail: info@adults.dev