What moves the number
Whether the design exists. A finished, well-built design file is the single biggest lever there is. Design and build together is a different project from build alone, and it is the first thing any honest quote asks about.
How much of the content you can edit yourself. Pages are cheap to build and expensive to own; collections cost more up front and hand you the keys. That choice moves a project more than page count does.
The number of genuinely distinct layouts, not the number of pages. Twelve pages built from four layouts is a smaller job than six pages that each look nothing like the others.
Anything with state: forms that route somewhere, a booking flow, a search that filters, a second language. Each of these is a system rather than a screen, and each moves the band.
Whether the content exists. Waiting on copy is the most common reason a fixed scope turns into an open one, and it is a cost even when nobody invoices for it.
Roughly what each band buys
$1k–$3k: one page doing one job. A landing page with a handful of sections, one call to action, built from a design that already exists. Zero is this shape: a single screen, one promise, one button, and the hover state designed as carefully as the rest.
$3k–$7k: a marketing site. Several pages, one or two collections you can edit yourself, a form that routes somewhere real. Dein Sitz sits here: a German-language site selling a Munich business address, with a booking flow and the terms in the hero.
$7k–$15k: a bigger site, or a smaller one designed from scratch. More layouts, several collections, more than one flow, sometimes a second language. This is also where design-and-build lands when there is no file to start from.
$15k+: something with a system behind it. Accounts, per-user data, real integrations, a marketplace, or an ongoing build rather than a delivery. At this point the question stops being what the site costs and becomes what it has to do.
Where budgets actually get eaten
Almost never by the thing that was quoted. It is the page added after the design was signed off, the second decision-maker who appears in week three, and the content that was going to be ready.
None of those are unreasonable: projects change. The problem is discovering them one at a time, which is how a clear number becomes a running total.
The cheap fix is naming the scope precisely enough that a change is visible as a change. If adding a page is obviously adding a page, everyone can decide whether it is worth it. If the scope was vague, nobody can tell the difference between a change and a misunderstanding.