Three quotes for what you described as the same website, and the gap between the lowest and the highest is wider than the lowest quote. Nothing in the three documents explains the gap. They describe roughly the same site, list roughly the same pages, and not one of them tells you which part of your project is the expensive part.
That is worth sorting out before you compare numbers, because the thing creating the spread is rarely the thing the quote is itemised against.
What actually moves a website price
A website price is set by how much unresolved work the project still contains. Not by how large the finished site looks. Four things account for most of the variation, and page count is not among them.
How much of your content already exists in a usable form
This is the largest single driver and it is not close.
“Usable” is carrying weight in that sentence. Copy that exists as a company profile PDF from six years ago is not usable content. Neither is a folder of site photos with no dates, no scope and no client permission attached. Neither is “our team can send you something next week”, which in practice means the project is now paying for someone to interview your engineers and write it up.
A site where the structure, the copy and the images are already decided is a design and build job. A site where they are not is a research job with a design and build job attached. The second one costs more because it takes longer, involves more of your people, and produces most of its value before anything is designed.
Across 54 Singapore B2B sites we reviewed in August 2026, 19 published no project record, portfolio or case studies at all. Those companies have all done the work. None of it existed in a form anyone could put on a page. For a business in that position, the first phase of a rebuild is spent creating the raw material, and that phase is in the price whether or not the quote names it.
How many people have to approve it
Rounds are the unit of cost in this work, not pages.
Harvard Business Review’s research on B2B buying is blunt about the direction of travel: unilateral decision makers are now rare, and most purchases are made by groups. Buying a website works the same way. A project with one person who can say yes moves at one speed. A project where every screen has to clear a managing director, an operations head, a compliance officer and a founder who is offshore for three weeks moves at another, and the difference is not effort, it is calendar.
Nobody prices approvers as a line item. It arrives anyway, as extra rounds, as re-work on a phase that was already signed off, and as a launch date that quietly slides.
How many systems it has to talk to
A booking flow, a customer portal, a product or project database, a payment gateway, a CRM someone in sales actually uses, an ERP, a language switcher with genuine translated content behind it.
Each of these is a small project wearing the same word as a contact form. It has its own testing, its own failure modes, its own credentials to obtain from a third party who is slow to answer, and its own maintenance afterwards. This is where fixed-price quotes usually go wrong, because the integration looked like one line in the brief.
How much technical material has to be gathered from scratch
This is the industrial version of the content problem and it deserves its own heading. Certification scope, tendering grades, capacity figures, tolerances, vessel types, method statements, equipment lists, approvals with their real expiry dates.
All of it exists inside the business, usually inside two or three people who are busy. Getting it out, checking it, and writing it so a buyer can act on it is the most valuable part of an industrial web project and the part most quotes are silent about. When we rebuilt Hong Hock Global, a water infrastructure contractor whose work is awarded on prequalification rather than persuasion, the expensive part was never the layout. It was working out what a PUB officer needs to see, and in what order, and then getting it stated plainly.
What doesn’t move a website price?
Page count. Five pages and fifteen pages are often close in cost, provided the extra pages are variations on a structure that has already been designed and the content exists. A page is cheap. A decision about what a page is for is not.
The number of colours, typefaces or animations. These are outputs of a design decision, not inputs to a price. A studio charging more for a second typeface is charging for something else and labelling it badly.
“Make it pop.” This is not scope. It is feedback with no decision behind it, and it costs money only in the sense that it generates rounds. The way to avoid paying for it is to say what is not working, rather than what should be added.
The platform, most of the time. WordPress, Webflow, Framer or something custom changes who can maintain the site, what it costs to run and how it behaves in five years. Those are real considerations and worth deciding properly. They rarely explain a large gap between two quotes for the same brief.
Responsive layouts. Any quote listing these as a feature is padding the deliverables list. They have been table stakes for a decade.
What genuinely adds cost that nobody warns you about
Performance targets on a heavy site. Google’s guidance puts a good Largest Contentful Paint at 2.5 seconds or under at the seventy-fifth percentile of real visits. Hitting that on a site carrying large equipment photography, drone footage and a project archive is engineering work, not a checkbox, and it is worth agreeing as a target before the design rather than discovering it after launch.
Photography and art direction. Almost always underestimated, almost always the highest-return spend on an industrial site.
A second language with real content behind it. Translation is the cheap part. Maintaining two versions of a site that keeps changing is the expensive part.
Migrating an archive. A decade of news posts or product pages with inbound links attached is a project of its own, and skipping it quietly costs search visibility that took years to build.
How the money is staged
Smaller projects run 50% to start and 50% at launch. From S$8,000 we use three milestones: 25% on signing, 35% on design sign-off and 40% on launch, with 30 day payment windows on the last two so a finance department is not the bottleneck. Handover includes 30 days of support for bug fixes, content edits and questions, extended to 45 days on larger builds. Those terms are published rather than negotiated per client, which is deliberate. Payment structure is one of the few parts of a web project a buyer can compare like for like.
How to bring your own number down
Three moves, all of which remove rounds rather than reduce the work.
Do the content inventory yourself before you ask anyone for a price. List the services, the projects you can name, the projects you can describe without naming, the certifications with their real expiry dates, and the photographs you own the rights to. Whatever is missing from that list is the real scope of the project, and knowing it in advance turns a vague quote into a specific one.
Nominate one person who can approve on behalf of the business. Not one person who collects opinions. One person who can end a discussion.
Decide the integrations before anyone quotes. A booking system added in week six is not a change request, it is a different project.
Every figure is a starting point
Projects start from S$4,000 for a refresh where brand and copy are already in place, and from S$8,000 for a site built from scratch where structure and copywriting are part of the work. Industrial and corporate builds carrying certification content, technical explanation and integrations start from S$25,000.
Those are starting figures, not prices, and any studio that quotes you a fixed number against a two-paragraph brief is either padding it or planning to cut something later. The useful question to ask is not what the number is. It is which of the four drivers above applies to you, and what the quote assumes about each one.