How long a website takes, week by week
Week one is decisions, not screens. Who the site is for, the one action it has to produce, what pages exist, and which of them are collections rather than static pages. We also find out what content actually exists today, because that is the single biggest predictor of whether week three holds.
Nothing gets drawn until those answers are written down. A design made while the offer is still moving is decoration with a deadline, it gets redone, and the redo is what eats the schedule.
Week two is design. Every page drawn at desktop, tablet and phone, with the states included: hover, empty, error, loading. States are where “nearly finished” designs go to die, so they are part of the deliverable rather than something the build invents later.
Week three is build, content load, QA and launch. The CMS collections get their fields, the forms get their routing, the pages get tested at the three widths they were drawn at, and the site goes live. This is also where a well-built design file earns back the week it cost: what transfers from Figma and what gets rebuilt is decided long before anyone opens the Designer.
One rule holds the whole thing together: one round of changes per stage, and then the stage closes. Not because feedback is unwelcome, because feedback that arrives after the next stage started is a rebuild wearing the word “tweak”.
When the date isn't yours
Some projects have a deadline that belongs to something outside the project: a funding round, a trade show, a season, a programme that runs on a calendar. The Bridgeman Awards site is that shape: an awards programme where entries open and winners get published on dates nobody on the project controls.
When the date is external, it should be the first thing in the brief, not a detail that surfaces in week two. It changes what we recommend: fewer pages built properly beats more pages built to a date.
What makes it four
Content that arrives late. This is the cause in most cases that slip, and it is almost never the client being slow, it is that nobody owned the copy. Decide the owner in week one.
Fonts that turn out not to be licensed. Discovered in week three, this stops a launch cold. It takes ten minutes to check in week one.
A second decision-maker who appears late. If someone can veto the design, they should see it in week two, not after it is built.
Scope added after sign-off. A new page in week three is a new week. That is not a penalty, it is arithmetic, and it is a fine trade to make knowingly. It also usually moves the budget band, which is a better conversation to have out loud than to discover on the final invoice.
What we need from you, and when
Three weeks is not three weeks of us working alone. It is a schedule with your name on four specific days, and every project that slipped, slipped on one of them.
Day one: the answers in a workable brief. Who the site is for, the one action, what content exists, what has to connect to something else, and the one name that signs off. Half-formed is fine. Absent is what turns week one into week two.
End of week one: sign off the structure. Not a design, a list — the pages, which of them are collections, and what each one has to answer. Approving this takes about twenty minutes and it is the cheapest approval in the project, because everything downstream is built on it.
End of week two: one round of design feedback, from everyone who has an opinion, in one document. One round, consolidated, is the rule that makes the third week possible. Feedback that arrives in pieces over four days costs more than the feedback itself is worth.
During week three: the content, in its final form, and a decision on where the forms go. An email address, a CRM, a Slack channel, anything real. A form routed to nowhere is the most common thing found in a post-launch audit.
Everything else is ours. But those four are genuinely blocking, and naming them on day one is the difference between three weeks and the version where nobody quite knows who is waiting for whom.
When three weeks is the wrong answer
It is a scoping discipline, not a claim about every project, and there are shapes it does not fit.
A brand that does not exist yet. If the logo, the voice and the palette are being decided at the same time as the pages, the site is downstream of a branding project and should be scheduled after it, not around it.
Content that has to be created rather than collected. Twelve case studies to be written, a photo shoot to be arranged, forty products to be described. That is a content project with a website at the end, and its timeline belongs to whoever is writing.
Anything with state behind it. Accounts, per-user data, a live integration that has to answer while the visitor waits. Those are builds, not layouts, and pretending otherwise is how a three-week promise becomes a four-month apology.
A translation into a second language, if the copy is not already written in both. The build carries it easily; the words are the schedule.
And a migration with real traffic behind it. The build fits the three weeks; the redirect map and the recrawl do not care about your calendar.
In each of those the honest answer is a longer number, given before the work starts. Fewer pages built properly beats more pages built to a date, and a date invented to win the project is the most expensive thing in any proposal.
Questions we get asked
How long does it take to build a website?
For a marketing site with a defined scope, three weeks from brief to launch: one week of decisions, one of design, one of build and QA. A larger site designed from scratch runs four to eight. Anything with user accounts or a live integration behind it is a build rather than a site, and is measured in months.
Why do some agencies take three months for the same site?
Almost never because the work is bigger. It is because the scope kept moving, feedback arrived one message at a time, or the copy was still being written while the pages were being drawn. Time in a web project is mostly spent waiting for decisions, not making screens.
What is the fastest a website can be launched?
A single well-scoped landing page from an existing design can go live in a few days. That is a real answer and a narrow one: it assumes the copy exists, the design is finished, and nothing has to connect to another system. Remove any one of those and the number stops being days.
What usually makes a website project run late?
Four things, in order: content that arrives late because nobody owned it, fonts that turn out not to be licensed for the web, a second decision-maker who appears after the design is approved, and scope added after sign-off. All four are visible in week one if anyone asks.