Building Got Cheap. Owning Didn't.

The number that stopped mattering

Two years ago, if you spread the cost of a small business website over five years, the build was most of it. Everything after launch was hosting, a domain, and someone’s afternoon here and there.

That ratio has broken. Not because running a website got dearer — though in places it has — but because building one got dramatically cheaper. I can now produce in a week what used to take a month, and I’m not unusual in that.

So the build quote, the number everyone negotiates hardest, is shrinking towards being the smallest line on the page. Meanwhile the running total sits there unexamined.

This is my attempt to put a real number on the second one.

The eleven lines nobody quotes you for

A build quote covers design and construction. Here’s what it doesn’t cover, and what you’ll pay for over five years whether you planned to or not.

1. Hosting and bandwidth — per published site, usually with a traffic ceiling and a charge above it.
2. The workspace or team subscription — separate from hosting on most platforms. This is the one people miss entirely, because the pricing page leads with the hosting number.
3. Per-seat fees — every person who needs to edit. Scales with your team, not your site.
4. Metered AI usage — new, growing, and covered properly below.
5. CMS and content ceilings — fine until they aren’t.
6. Add-ons replacing removed native features — authentication, automation, search, form handling. See When the Platform Removes Something You Depend On.
7. Domain, DNS and business email.
8. Analytics and consent tooling.
9. Security patching and dependency maintenance — either included in your platform fee or your responsibility. Never absent.
10. Change throughput — the retainer, or the internal hours, or the opportunity cost of changes that never happen because they’re too much hassle. Usually the largest single line and almost never on anyone’s spreadsheet.
11. Exit cost — what leaving would cost, amortised as the risk it is.

Metered AI, and why it’s different

This deserves its own section because it’s genuinely new and it behaves unlike the other lines.

In May 2026 Webflow restructured its pricing and bundled AI credits into every Workspace tier: 200 on Starter, 300 on Core, 400 on Growth. Enforcement began on 29 June 2026, and once you’re out, more credits cost $20 per 2,000.

I want to be fair about this, because the reflexive complaint is that it’s expensive, and that isn’t quite the problem. The problem is that a credit isn’t a published unit.

You can’t see which model is running. You can’t see the token accounting. You can’t work out, before you’ve had a month of usage, whether 300 credits is generous or nothing. Two teams doing superficially similar work can burn wildly different amounts and neither can explain why. That’s a forecasting problem, and forecasting problems are worse than price problems because you can’t plan around them.

Now the honest comparison. On a coded stack you’d pay a model provider directly, at published per-token rates, and the meter is completely legible — you can see exactly what a task cost and predict the next one. That’s a genuine advantage. But you carry the integration work, the key management, and the responsibility when something loops and spends. Nobody hands you that for free either.

Neither approach is cheap. One is forecastable. Which matters more depends on whether you’re the sort of business that can absorb a surprise line on an invoice.

Three worked examples

Replace my assumptions with yours — that’s the point of showing the workings. I’ve used USD for platform fees because that’s how they’re billed, and left your own labour costs as variables.

A. The eight-page professional services site

A consultancy, a clinic, a chambers. Brochure pages, a contact form, a handful of updates a quarter. One person occasionally changes some copy.

On a managed platform: a Premium site plan at $25/month billed annually, a Core workspace, one seat. Call it roughly $40–50/month all in, plus £300 for the two or three afternoons a year someone needs help.

On a coded stack: hosting is nearly free at this scale, the CMS free tier covers it, and you’d spend much less on the build. But every content change routes through someone technical, and you’re carrying dependency updates on a site that generates a few enquiries a month.

The builder wins this comfortably, and I’d say so to anyone. The managed platform is absorbing security patching, backups, SSL and uptime for less than the cost of one billable hour a month. That’s a good trade and the ownership argument is mostly noise at this scale.

B. The forty-page content-led SME

A blog, case studies, a gated resource, campaign pages. Three people edit. Marketing ships something most weeks.

Now the seat count starts driving the bill — full workspace seats run $39/month each — and the workspace tier matters more than the site plan. You’re at $150/month before anyone does any work.

But the bigger number isn’t the subscription. It’s change throughput. If your marketing lead can build a campaign page herself on a Tuesday afternoon, that’s worth more over five years than any hosting saving. If every change is a ticket to an agency, you’re paying for that too x however many changes, and you’re paying again in the changes that don’t happen because nobody could face raising the ticket.

This one’s genuinely close, and it turns entirely on who edits it. Not on the technology.

C. The site that’s becoming a product

Logins, member areas, tiered pricing, integrations, data you’re responsible for.

Here the managed platform starts working against you. Webflow retired its native Logic automation in June 2025 and User Accounts in January 2026, pointing users to Memberstack, Outseta, Zapier and Make. Those are decent products, but you’re now assembling and paying for a stack, and the pieces you depend on most are the ones your platform decided not to own.

You’ll also start meeting architectural limits rather than effort limits — a 50,000 character cap on the custom code block, no control over your own build pipeline, hosting you can’t configure.

Coded or headless usually wins here, and the crossover point arrives earlier than most owners expect — typically the moment “website” becomes “the thing customers log into”.

Which decisions are actually expensive

If you rank the levers by how much they move the five-year total:

1. How many people need edit access. Dominates everything. Seats scale linearly and never go down.
2. Whether marketing has to wait for a developer. Not a line item, which is exactly why it’s underestimated. Measure the changes that don’t get made.
3. Whether anything revenue-critical sits on a bundled feature you don’t control.
4. How much genuine architectural headroom you need. Not how much you’d like.
5. Hosting. Last. Almost always trivial. And overwhelmingly where the sales conversation goes.

I’ve sat in a lot of meetings where an hour went on comparing $18 against $25 a month while the actual difference — whether the client’s marketing manager would be able to publish anything herself — went undiscussed.

Six questions, one rough number

Work through these and you’ll have a five-year band good enough to make a decision with.

1. How many people need to publish changes? (× seat cost × 60 months)
2. How many changes a month happen, and who makes them? (× your hourly cost, or zero if in-house and self-sufficient)
3. Does anything you charge money for depend on a feature bundled into your platform? (If yes, add the cost of its standalone replacement)
4. What’s your realistic traffic ceiling, and what happens above it?
5. Who patches, backs up and monitors it — and if that’s you, what’s your time worth?
6. What would leaving cost? (Find out here, then amortise it across five years)

What I’d actually tell you

Get a five-year figure before you get a build quote. If you already have a site, get one before you agree to a rebuild.

Most of the time the number is unremarkable and you carry on. Occasionally it’s four times what you assumed, and you find that out now rather than in year three.

I sketched the twelve-month version of this argument in Visual Editors vs. Vibe Coding, which looks at the same question from the point of view of where your developer time lands. This is the longer, more financial version.

Platform pricing verified 13 August 2026. Webflow bills in USD; convert at whatever the pound is doing this week.

→ Back to Rented Ground
→ Next: The Exit Test