There's a pattern we meet so often in diagnostics that we've started calling it by name: the twice-built website.

It goes like this. A business pays for a website. It launches, it looks good, everyone's proud of it. Then it sits there. Months pass, the enquiries don't come, and the quiet conclusion forms that websites just don't work for businesses like ours. Eventually someone credible looks under the bonnet and finds the real problem: the site was never built for a single search that anyone actually makes. So it gets rebuilt. Same business, same product, second invoice.

Sometimes there's a third. The ads get switched on, the clicks arrive, and the pages were never designed to convert a paid visitor, no message match, no tracking, no path. Rebuild number three.

Here's the thing worth sitting with: none of those websites were badly designed. They were designed in the wrong order.

The order that causes it

The default sequence runs look first, findability later. Choose a designer, pick a style, write the pages around the company menu, launch, and only then ask how people will find it. By that point the answer is baked in: they mostly won't, because the searches that matter were never in the brief.

Search demand isn't a mystery you discover after launch. It's measurable before you spend a dollar. What your market types into Google, what they ask ChatGPT and Gemini, which competitor owns which answer, and where the gap is, all of it can be mapped in about a week. Which means the pages a site needs, and the language they need to speak, can be known before design begins.

When one of our diagnostics ran on a three-decade-old tour business last winter, the map said the site needed 46 pages, told us which searches each one had to win, and showed the audience the design had to serve. The palette, the photography, the tone, all of it followed the demand. That site produced a year's worth of enquiries in its first month. Not because the design was prettier than the old one, though it was, but because every page opened onto a search real travellers make.

The order that fixes it

Demand first. Measure what your market searches and what a customer is worth to you, and the arithmetic of the build is on the table before you commit. If the demand isn't there, you just saved yourself a website.

One language second. The brand, the copy, the pages and the ads built together so everything says the same thing in the same voice. This matters more now than it ever has: AI assistants cite businesses with a clear point of view and absorb generic copy without a mention. Sounding like everyone else has become a visibility problem.

Measurement always. Every form tracked, every call counted, results scored against a published projection monthly. A website you can't measure is a website you'll be sold twice.

And build it ads-ready even if ads come later, because a paid click landing on a page that wasn't built for it is the most expensive kind of traffic there is.

The one-payment checklist

• Before design starts, you can name the exact searches the site is built to win, and the data behind them.

• Every important page answers one of those searches, in your voice, with your proof on it.

• The site is structured so AI assistants can quote it, clear answers, real schema, a point of view.

• Every enquiry path is tracked from day one, forms and phone calls both.

• There's a projection on paper, and a monthly habit of scoring actuals against it.

• You own the site, the accounts and the data, whoever builds it.

If a quote in front of you can tick that list, take it, whoever it's from. If it can't, you're probably looking at the first half of a twice-built website.

We wrote the full market guide to what websites cost in NZ and Australia, bands, inclusions and the questions to ask any provider. If you'd rather start with the measurement, that's what Sphynx 2.0 is for: your demand, your competitors and your gap, mapped in about a week for NZD $489. And if you want to see what demand-first design looks like in practice, that's how we build websites.