Trendo.mx

The agency's problem was not quality. It was that its research lived in PDFs behind a login, where neither search engines nor language models can read it, and the site did not rank for its own subject. Search the topic this agency has spent years owning, and you find global consultancies and media agencies instead.

The audit put overall SEO health at roughly 5 out of 10, and the shape of that score is the interesting part: strong brand, weak technical base. Around ten indexed pages. No meta descriptions on any page audited. A flagship landing page with no H1 and roughly 180 words on it. A sitemap still listing sections last touched in 2020, and a blog that stopped in late 2024.

The figure that reframed the whole engagement did not come from any analytics tool. It came from the client, buried in a quote from months earlier: the agency records roughly forty thousand PDF downloads a year, behind a login, with no usable email capture attached.

That changes the diagnosis from a traffic problem into a capture problem. Forty thousand people a year explicitly ask for this agency's research, and the agency keeps almost none of them. Read alongside the finding that branded and organic traffic move as one line, the implication is that essentially all of that demand arrives through rented channels — social platforms — and switches off the day an algorithm changes.

Replacing the login with an email gate is not a conversion tweak. It is the difference between an audience the agency owns and one it borrows.

Auditing each funnel stage against that persona — does this move her one step forward? — produced a harsher score than the SEO audit did, and a more useful one. The site is not underperforming evenly; it is missing its mouth.

Five leaks, in order of what they cost. The expensive one is at the entrance: someone searching the topic never encounters this agency at all, and leaves with a global consultancy instead. Second, an enormous social audience with no destination that converts. Third, the people who do arrive — already knowing the brand — cannot assemble an internal case, and stall. Fourth, the friction of a login on the reports plus no easy way to say "let's talk." Fifth, nothing continuous to keep her engaged, so no repeat business and no referrals.

Partway through the design phase a homepage was built from a visual reference — and it copied the reference's structure along with its look. The result linked out to a video platform, dissolved Trends and Reports as destinations, and put selling ahead of attracting. Every one of those is a direct violation of the architecture that had already been agreed and written down.

It was rebuilt. What came out of it is now a standing rule on the project: a visual reference contributes art direction, never information architecture. If a reference implies a different structure, the funnel wins. Writing that down as a rule rather than a one-off correction is what keeps it from happening again on the next page, with a different reference and the same logic.

Decided: the content mission and the public tagline, phased scope with the redesign first and the content engine second, and a site scoped at nine unique pages plus three reusable templates. The 108 URLs of the existing site are inventoried and classified into keep, redirect, or retire, so that the migration preserves the referring domains rather than resetting them to zero — new design, existing equity.

Open, and honestly so: the platform recommendation has not been accepted by the client yet, and the alternative costing exists in case it is refused. Read access to the client's search console and analytics has been requested and not yet granted — which is the riskiest gap, because without knowing which URLs actually carry those backlinks, the redirect map gets prioritized blind. Several assets the plan depends on are still with the client: case permissions, testimonials, and the email tooling.

What I'm proud of: finding the forty-thousand-downloads figure in an old quotation rather than in an analytics dashboard, and letting it overturn a diagnosis I had already written. What I'd do differently: I built a homepage from a visual reference and let it import an architecture I had already argued against — I should have caught that in the first review, not the third.