Growth in a category where every claim has to be provable.
What We Did
01 . Client overview
The client
Lumen Harley Street is a private health assessment clinic at 17 Harley Street, in the middle of London’s medical district, and part of the UME Health group. It runs a tiered range of advanced health assessments, Lumen One, Lumen Two and Lumen Three, alongside corporate health assessments and add-on diagnostics including MRI, ultrasound, CT coronary imaging and FibroScan liver scanning.
Assessments are delivered by GMC-registered doctors, with Dr Garry Savin leading the clinic. Pulse Digital Health holds search, content, structured data and the technical health of the site.
At the start of the period the site averaged position 30.9, which is page 3, and earned roughly 8,000 search impressions and 75 clicks a week. Almost all of the demand it was chasing sat on generic national terms it did not yet rank for.
Sector
Private healthcare. Preventative health assessment and diagnostic imaging, sold to self-paying patients and to companies, from a single address in the most concentrated private medical district in the UK.
Starting position
Average position 30.9 across the first 4 weeks of recorded data, roughly 8,000 search impressions and 75 clicks a week. Page 3 for the terms that mattered, and no ranking at all on most of them.
Remit
Grow non-brand visibility for the assessment range, make the clinic legible to search engines and to AI systems, and keep every claim on the site inside ASA and CAP rules.
02 . The challenge
The challenge
Private health assessment is a category where the demand is national, the vocabulary is generic and the regulator reads everything. Six constraints shaped what the work could and could not do.
A category with no local moat
Full body health check and full body scan are searched nationally and answered by imaging groups with multiple sites and far larger budgets. One clinic on one street competes for the same terms, so the content has to be better rather than nearer.
Every claim is regulated
This is YMYL content under ASA and CAP rules. Superlatives, implied outcomes and unsupported comparisons had to come off the site rather than go onto it, so the usual levers of persuasive copy were not available.
A brand name that is also a common noun
Lumen is a unit of light and an anatomical term. Brand search is contaminated by unrelated results, and AI systems return other entities of the same name, so brand queries could not be relied on to carry visibility.
A decision taken over weeks
An advanced health assessment is a considered purchase, often discussed with a partner before anyone picks up the phone. A click rarely converts on the same visit, so the content had to serve research as thoroughly as it served booking.
Clinical sign-off sets the pace
Nothing patient-facing publishes without a doctor reading it first. Publishing speed is governed by clinician availability rather than by the agency, so work had to be organised into batches the clinical team could realistically review.
An estate built by several hands
The site had grown through more than one build. Alt text carried marketing boilerplate, a physician portrait carried another physician’s name, internal links pointed at pages that no longer existed, and staging subdomains had been left indexed.
03 . Our strategy
Our strategy
The site did not need more pages. It needed the pages it already had to be understood, connected and provable. Five strands ran alongside each other.
01
Made the clinic legible as an entity, not just a website — A 10-page JSON-LD pack was written and deployed, carrying MedicalClinic, Organisation, WebSite, MedicalProcedure and Physician entities cross-linked through stable @id references, so Google resolves one connected graph rather than a set of unrelated fragments. Each assessment page carries its own MedicalProcedure entity pointing at the doctor who performs it, and the clinic entity carries the Harley Street address that anchors it locally. Two rounds of fixes were applied against the Rich Results Test until every item validated cleanly.
02
Mapped content to what patients search, not what the clinic sells — Search Console queries were bucketed by theme to find demand with no content behind it. The full body and whole body scan theme alone carries 240,111 impressions across 232 queries, and a page-one ranking for full body cancer screening in London was sitting at position 8.8 on a page with nothing written for it. The FibroScan liver article proved the model that answered this: one well-structured clinical explainer now holds an average position of 6.2 and earns 193 clicks, which is more than either of the two lower assessment pages it supports.
03
Rebuilt the internal link structure around clusters, not page count — 29 links were mapped across 6 clusters, core assessments, corporate, liver, heart, brain, and clinic and team. Each one specified the source page, the exact sentence it sits in, the verbatim anchor and the destination, so nothing was left to interpretation at implementation. A second pass covered blog to blog only and returned 9 links where the anchor text already existed in the body copy, which meant no new copy needed writing or clinical approval. Two broken internal links were found and reported in the process.
04
Fixed the image layer across the whole estate — 58 pages, 475 image instances and 101 unique images were audited against the live site rather than a crawl summary, with the images downloaded and looked at wherever the correct description was ambiguous. 16 images had no alt text at all, 44 were rewritten and the remainder were verified and kept. Marketing superlatives, repeated keyword boilerplate and CMS artefacts were stripped out, decorative icons were set to empty alt so screen readers skip them, and a physician portrait carrying the wrong doctor’s name was found and corrected.
05
Put every claim on a footing that would survive a challenge — Every award displayed on the site was traced back to its awarding body and year, and 16 entries were documented. Claims that could not be evidenced were flagged rather than repeated, including an award with no named awarding body and a total number of wins higher than the site could show. The homepage recognition section was rewritten to name the awarding bodies and the years instead of reaching for superlatives, which is both the compliant version and the more persuasive one.
04 . The results
The result
The work was judged on one thing: whether the right patients could find the clinic at the moment they were looking. Impressions have grown faster than clicks throughout, because the site now appears for a far wider set of terms than it started with, and average click-through rate falls as that happens.
The Numbers
Monthly impressions and clicks
Google Search Console, web search, all countries. The first month of recorded data is partial, and the partial final month is not plotted.
