A CRO whose science outran its website.

01. The situation

Ranomics does protein engineering: directed evolution, deep mutational scanning, AI binder design. The science is the hard part and the science was already working. The website was not carrying it. Its content had accumulated across years as a set of pages written for people who already knew what the company did, which is the single most common failure mode in this sector. A first-time reader arriving from a search result, a conference, or a partner introduction had to do the work of assembling the offer themselves.

02. The constraint that shaped everything

There were 36 pages of legacy content with existing search equity. The fastest way to destroy a life-science site is to rebuild it beautifully on new URLs and silently drop years of accumulated ranking. So URL preservation became a hard build constraint rather than a nice to have: every existing address had to survive the migration exactly, and the new information architecture had to be built around that, not the other way round.

That is the unglamorous half of a redesign, and it is the half that decides whether the project is a gain or a very expensive reset.

03. What was built

A 53 route static site on Astro, with 64 technical articles as a content collection. Static output because the pages are read far more often than they change, and a static site is both faster for a human and simpler for a crawler to parse.

Structured data was treated as part of the build rather than as an afterthought, through a single schema component applied across page types. This is the same work Bold48 sells as answer visibility: an engine that arrives with no context takes what the markup gives it, and most life-science sites give it almost nothing.

The commercial path was wired at the same time. Enquiries route through a lead capture endpoint into the CRM rather than into an inbox, and service pages carry a query parameter that pre-fills the enquiry form with the programme the reader was actually looking at, so context is not lost between the page and the conversation.

04. Where it stands, including where it falls short

Here is the part most case studies leave out. We ran our own AI-search scanner against ranomics.com on 27 July 2026 and published what it found, including the failures.

Organization and WebSite schema are present and valid. Every page has a single H1 and a real description. But there is no named author entity, so an engine cannot attribute the technical writing to a person, and the pages do not lead with a direct answer to the question they are named after. Those two gaps are why the composite sits at 40.2 rather than somewhere respectable, and they are the current work.

We publish this because a studio that will not measure its own work has no business measuring yours, and because a 40.2 we can explain is worth more than a 90 we cannot.

05. What is deliberately not on this page

No traffic chart. No ranking positions. No revenue attribution. No conversion lift. Those numbers are not here because they have not been measured cleanly, and a redesign that ships alongside content, outreach and product changes cannot honestly claim sole credit for a traffic line anyway. When we have a measurement we can defend, we will publish it here with its method attached.