Money is measured on the server
A confirmed purchase arrives by webhook, is idempotent and doesn't depend on the browser staying open. Any revenue figure measured in the browser is an estimate, and it ends up being used as if it weren't.
Everything you need to decide before anyone opens the editor
The kit that turns “we need a new website” into a brief that can be built: objective, personas, messages, journey, architecture, screen inventory, stack, measurement, design, risks and a phased plan.
Almost every website project starts with the visible part: references, colours, “I like the way this other company does it”. And almost all of them discover too late that the expensive part wasn't the design but the decisions nobody made: what the site exists for, which conversion takes priority over the rest, what happens to the old URLs, who maintains it a year from now.
This kit puts those decisions first. Twelve pieces: eleven for work and one for rules. They're used in order, because what's closed in one feeds the next. At the end there's no website; there's a brief that can be built without asking anything again.
Its little sister follows from it: one page, one job. If “services” are three lines of business, that's one page with three sections, not three subpages. Every subpage opened for convenience is a URL that has to be maintained, linked, ranked and, sooner or later, redirected.
Each frame decides one thing and one thing only. What's closed there isn't reopened in the next: if the temptation shows up, it's a sign the previous piece wasn't really closed.
What the site exists for, who it speaks to, what it says and which way that person comes through. It's half the work and the half that almost always gets dispatched in a half-hour meeting.
One primary objective—just one—the conversion that takes priority over all the others, the list of anti-objectives and the north-star metric it's measured by.
One persona per column, the three-layer company filter and the anti-persona: who the website has to repel, and by what mechanism.
Positioning, claim, differentiators, what proves each one and the voice guardrails—including the list of banned words.
Each stage with five rows: what's happening to them, what they're asking themselves, where it happens, what they need to see or feel, and what event measures it. With no event, the stage doesn't exist.
How many pages there are, how they hang off each other and what each one does. It's the first decision that gets expensive if it's done badly.
The map with the final URLs and, if there's a previous site, the redirect map from the inventory of old URLs.
One card per screen with its job, its persona and its status. Whatever has no card doesn't get built.
What it's built with, how it's measured, how it looks and in what order it goes up. This is where a project becomes maintainable or becomes a debt.
The choice for each layer—framework, content, data, payments, hosting, analytics—and, next to it, the rejected list with the reason for each.
One event per funnel step, the source of truth for each number and a single KPI per channel.
A palette with the role of each colour, two type families at most, rules of form and where each value comes from. If it isn't in the system, it isn't used on a page.
Three lists: what's decided, what's open and—marked separately—what's open and blocks everything else.
The build order and, in each phase, what's deliberately left out. What doesn't fit in phase 1 doesn't die: it waits.
The inventory isn't a list of page names. Each screen is written on a card with the same fields, and that card is the brief for that page.
| Field | What goes in it | Why |
|---|---|---|
| Route | The final URL, decided here and not during development. | Changing it later costs redirects and rankings. |
| Job | What this screen has to achieve, in one sentence with a verb. | If the verb is “show”, it's almost always a section of another page. |
| Who for | Which of the personas from frame 02, by name. | A page that serves everyone is written well by nobody. |
| Event | What fires when it does its job. | It's the condition for getting onto the map: without an event you can't tell whether it works. |
| Status | Written, awaiting copy, awaiting a decision or blocked. | It separates what's waiting on development from what's waiting on a person. |
The stack piece has two columns. In the first, what's chosen for each layer. In the second—the one almost nobody writes—what's been ruled out and why. Without that second column, six months from now someone will propose exactly what was already ruled out, and the whole discussion will be repeated because the reason isn't written down anywhere.
The principle behind it: master one thing thoroughly rather than renting five platforms by halves. Every service you adopt is another admin panel, another mental model and another data silo. Adopting has a cost that doesn't show up on the invoice.
They're decided before building, not after. An event that wasn't instrumented in time can't be reconstructed after the fact: that data simply doesn't exist.
A confirmed purchase arrives by webhook, is idempotent and doesn't depend on the browser staying open. Any revenue figure measured in the browser is an estimate, and it ends up being used as if it weren't.
Acquisition analytics and product analytics don't have to match: they measure different moments. Write down what each one measures and which one rules on each question, or every meeting will start by comparing two dashboards.
A client's real status is in your tables, not in a third-party dashboard. Everything else is reconciled against that. If a number matters for a decision, it has to be possible to query it without depending on an external service.
And one KPI per channel, not a dashboard of thirty metrics. Five numbers on one screen, or it doesn't get looked at. The rest lives in a monthly report that can actually be read in full.
Piece 10 has three lists, not two. What's decided, what's open, and what's open and blocking. That third list is the one looked at in every follow-up meeting, and it's usually short: two or three things.
A gap is filled with a fact or declared pending; it's never filled by making something up. Writing “[ pending ]” on a legal page is uncomfortable and honest. Writing a plausible identifier is comfortable and false, and nobody will ever look at it again.
PDF workbook, with the twelve pieces and their questions
The twelve pieces with their questions, the blank screen card and the closing checklist. To take to the session on paper.
Everything in the file is also on this page, open and without asking for anything. The email is so we can send you the version you can print and take to a session.
Done. Here it is.
Download the workbook as a PDFIf the file doesn't open, save this link: it will keep working.
You decide which cookies we use
With your permission, we measure, in aggregate, how quether.com is used and remember your visit to show you Quether ads on Facebook and Instagram. You can accept everything, reject everything or choose. More detail in our cookie policy.