Web development cost. Priced by scope, not pages.
Nobody can quote a website by page count, which is why every "average cost" article is useless. Here is what actually moves the number, the three bands most projects fall into, and how to tell which one you are in before you talk to anyone.
On this page
- 01Three bands cover almost everything: a marketing site, a CMS-driven site the team runs themselves, and a web application. They are different products, not sizes of the same one.
- 02Page count is a weak predictor. Twenty pages from one template is cheaper than five pages that each need their own design and logic.
- 03Integrations are the biggest silent multiplier. Every system the site must talk to adds discovery, error handling and a support surface.
- 04Content is the most common reason projects slip. Design and build finish; copy does not. Price it, or the schedule is fiction.
- 05Budget for the year, not the launch. Hosting, maintenance and iteration are the cost of owning a site, and they are where a cheap build gets expensive.
Which one are you actually buying?
Marketing site
A designed site that presents the business well and converts. Templated sections, a small number of unique layouts, forms and analytics.
Bought by: companies whose site is a shop window, not the product.
CMS-driven site
Everything above, plus a content model your team edits without a developer: structured content types, editorial workflow, multi-author, sometimes multi-language.
Bought by: teams publishing regularly, where the bottleneck is developer time.
Web application
Accounts, permissions, data the user creates and owns, business logic, third-party systems, and a real testing burden.
Bought by: businesses where the software is the service.
The jump from band two to band three is the one that surprises people. It is not incremental. Authentication, roles, state and data ownership change the engineering, the QA and the ongoing cost together. If your brief contains the word "login", you are in band three, whatever the page count says.
Seven things that actually change the price.
Unique layouts, not pages. Count the templates. A site with six templates and sixty pages is a six-template project.
Integrations. CRM, ERP, payment, booking, marketing automation. Each one is discovery, mapping, error states and a thing that can break at 2am.
Content model depth. A blog is cheap. Structured content with relationships, taxonomies and editorial permissions is a build in its own right.
Design origin. Adapting a system you already own costs less than creating one. If brand work is in scope, that is a separate line, and it is worth it when the brand is the differentiator.
Migration and redirects. Moving an existing site with rankings means a full URL map and redirect testing. Skipping it is how sites lose traffic on launch day.
Accessibility and compliance. Meeting WCAG properly is design and engineering work, not a checkbox at the end. In many sectors it is also a legal requirement.
Performance targets. "Fast" is free. Hitting Core Web Vitals at the 75th percentile on real field data, on a content-heavy site, is deliberate work.
Why this page has no price table.
Because a table would be a guess dressed as a fact, and you would find out it was a guess after you had planned around it.
The honest version: the three bands above differ by roughly an order of magnitude between the first and the third, and within each band the drivers in section 02 can double the number. That is not evasion, it is the actual shape of the market, and any agency printing a single figure is either quoting one narrow product or hoping you will not ask.
What we will do is put a real number on your scope quickly. Tell us which band you are in, how many templates, which systems it must talk to, and who writes the content. That is normally enough to give a range in one conversation and a firm quote after a short discovery.
Four ways to cut cost without regret.
Cut templates, not quality. Fewer, better layouts beat many mediocre ones, and they cost less to build and maintain.
Have the content ready. The single largest avoidable cost is a build waiting on copy. Write first, or budget for someone who will.
Phase the integrations. Launch with the one that earns money. Add the rest once you know how people actually use the site.
Do not buy band three by accident. Plenty of "portals" are a well-structured CMS site plus an email. Test whether you truly need accounts before you pay for them.
Six answers.
How much does a website cost in 2026?
It depends which of the three products in section 01 you are buying, and they differ by roughly an order of magnitude. A marketing site, a CMS-driven site and a web application are not sizes of the same thing. Identify your band first; the number follows from that plus the drivers in section 02.
Why do agency quotes vary so much for the same brief?
Usually because they are not quoting the same brief. One has assumed your content is ready and one has not. One includes accessibility, testing and a redirect map; one will raise those as change requests later. Compare what is excluded, not just the totals, and ask each quote who writes the copy.
Is a template cheaper than a custom build?
Upfront, yes, and for many businesses it is the right call. It stops being cheaper when you fight the template: every hour spent forcing it to do something it was not built for erodes the saving, and heavily customised templates are often harder to maintain than a clean custom build.
What should we budget after launch?
Hosting, dependency and security updates, and iteration. A site that never changes after launch is a site nobody is using to sell. Treat maintenance as an operating cost of the channel, the way you would with paid media. See website maintenance services.
Does the technology choice change the cost?
Less than people expect for the build, more than they expect for the decade after. Scope drives the build cost. The stack drives who can maintain it, how easily you hire, and whether you are locked in. Choose for the second reason. Our Webflow vs WordPress comparison works through one common version of this decision.
How long does a build take?
Band one runs in weeks, band two in a couple of months, band three in a quarter or more, and every one of those assumes decisions and content arrive on time. In our experience the schedule is set by the client's slowest approver far more often than by engineering.
Describe the scope. Get a range today.
Tell us your band, roughly how many templates, what it has to integrate with, and who writes the content. That is enough for a real range in one conversation, not a brochure.
Published .
