Forty three pages, in fourteen days.

Nobody could answer where a single one had come from.

Every advice firm in the country has the same three pages. Mortgages. Insurance. KiwiSaver. It is how the industry describes itself, so it is how the websites get built.

Nobody searches for insurance. They search for the thing they are worried about.

Someone working out what happens to a business partner's shareholding is not running the same search as a first home buyer trying to size a deposit. Same category, different person, different words, different page.

So the firm has nineteen services, not three. Nineteen pages, nineteen guides, nineteen forms, nineteen tables, nineteen pipelines, nineteen email sequences. Built once as a structure and generated, rather than made nineteen times.

Eight core pages, nineteen sub-service pages, two calculators and eleven long-form articles. Bespoke copy on every one. No category page pretending to serve four different audiences at once.

Self-referencing canonicals on all forty three, after a bug was found pointing every page at the homepage. Full postal, geographic and financial services markup. Indexed within days.

Every sub-service page carries two ways in. An enquiry form for someone ready to talk, and a gated guide for someone who is not. Both write to their own table, both carry the full attribution, and both trigger a different email sequence.

The guide is the qualifier. It does the work an adviser used to do on a first call.

Consent is captured at the point of submission and gates everything downstream. Nothing enters an email sequence without it.

Long-form, written to search intent, and deliberately more useful than the competition. The questions people actually type are not the questions the industry publishes answers to.

Conversion events fire only once the record is confirmed saved. A failed submission reports nothing to any advertising platform. That single ordering decision is the difference between a conversion count you can trust and one you cannot.

Three seconds to the CRM. Six to the email platform. The sync trigger lives in the database rather than the front end, because a front-end configuration value is one bad deploy away from silently disconnecting every lead the business receives.

Every integration was verified by reading the record back out of the receiving system rather than trusting a success response. Roughly a third of build steps initially reported success on something that had not landed.

It is an input to the page copy, the keyword list and the tracking payload.

No rates, percentages, dollar figures or thresholds appear anywhere across any channel. No projections, no guarantees, no urgency, no testimonial implying a financial outcome.

The tracking was tested to confirm no financial figure ever leaves the browser. A test submission carrying a loan amount was traced end to end: the figure appears in the database and in no outbound request to any advertising platform.

The financial services category was declared at campaign creation rather than added later, and the New Zealand advertiser verification was completed against the licence holder rather than the trading entity. Both are one-way decisions that cannot be corrected afterwards.

A disabled ad account that everybody blamed on velocity. The account had been shut down after a batch of ads went up quickly, and the obvious reading was that speed triggered it. The real cause was that the financial services category had never been declared on the campaign, and that setting cannot be changed after creation. We rebuilt the structure with it declared rather than appealing and repeating the fault.

Leads reaching the database and stopping, because a sync had been built, tested and never actually connected, while the code marked every record as delivered regardless. A CRM configured to treat two people sharing a household phone number as the same person, overwriting the first. Both advertising platforms optimising toward an event that visitors to those pages never performed. Two credential outages found by audit rather than alerting, one of which had been silent for two months.

Every one of those was found by comparing what the platform reported against what the database actually contained.

Structured interviews with both directors, then a content programme built around them rather than the firm. One came off the tools after three reconstructions and retrained while sitting on his own accident claim. The other was one of the youngest bank managers in the country and left when the bank was sold.

Both are camera shy and at their desks all day. So every shot in the first month is over the shoulder, from behind, side-on, hands only, or has nobody in it at all. Not one piece to camera. The whole week is captured in a single ninety minute stills shoot.

They asked to know which of their marketing was actually working.

The site answers the questions they used to answer on the phone. The guides qualify before anyone picks up. The pipelines catch enquiries without anyone watching an inbox. The flows follow up on their own. The tracking connects a click to a conversation.

The measure of this work is not impressions. It is whether an adviser can look at a lead and know where it came from.