The best conversion rate we have ever measured, on a two-page website
An 11.76% click-through rate — the highest of any site we run — on a website with two pages. We built the content system, and made the clinical sign-off a build failure rather than a promise.
The numbers
A conversion rate with nothing to convert
Compassionate Space Counselling is the private practice of Jane Hartland, a qualified person-centred counsellor. When we looked at the Search Console data it had the best click-through rate of any site we manage — 11.76%, where a good rate is 3–5% — and its average position had improved from 43 to 15 in four weeks.
When people saw it, they clicked. The problem was that almost nobody saw it, because it was a two-page website. It appeared for 22 search terms in an entire quarter. Its privacy policy was pulling as many clicks as its homepage.
There was nothing wrong with the site. There simply was not enough of it to rank for anything.
The reframe: the site is not there to win Google
Before writing anything we looked at what actually dominates counselling searches. It is directories, overwhelmingly — the BACP Therapist Directory alone lists over 18,000 therapists, alongside Counselling Directory and Psychology Today.
An individual therapist's website essentially never outranks those for a generic term, and pretending otherwise would have meant selling a year of content aimed at a fight that cannot be won. People search a directory, filter by issue and location, then click through to individuals.
So the site's job is not to win the search. It is to convert somebody who has already found Jane in a directory and is deciding whether to trust her. That changed what we built, and it meant our first recommendation to the client was to spend money somewhere other than with us — on the directory listings.
Governance the build enforces, not the calendar
Mental health content is what Google classifies as YMYL — your money or your life — and it is assessed on genuine expertise. It also carries a real professional risk that has nothing to do with search: it is Jane's professional standing on the byline, not ours. One article she has not actually read, making a claim she would not make, is a problem for her career.
Most agencies handle that with a promise. We handled it in the schema.
An article carries `approvedBy` and `approvedDate` fields. The moment a draft is marked for publishing without them, the site fails to build — with an error message that explains why. Not a warning, not a linting rule somebody can skip. The deployment stops. We tested it both ways before shipping.
The topic list is a closed set of the nine areas Jane approved. The areas she and her supervisor have not agreed — self-harm, suicidal thoughts, eating disorders — are absent by design and cannot be added by writing an article about them. Crisis signposting lives in the page layout rather than in article text, so it cannot be edited out of an individual post.
Two independent gates: a human one where Jane reads and approves, and a mechanical one that refuses to build if she has not.
What we shipped
An articles platform on Astro with the collection, routes, layout and structured data — marked up as an article by a named counsellor with her credentials, deliberately not as medical or treatment content.
A 52-week publishing schedule, every Monday at 09:00, with British Summer Time and GMT resolved correctly across the year rather than drifting by an hour in October.
Six articles drafted for review, an image house style derived from the site's own photography, and continuous integration that did not previously exist — so the governance gate runs on the way to production rather than only on a developer's machine.
Honestly
The articles are drafted, not published. They are waiting on Jane to read and sign off, which is the entire point — a byline that says a named counsellor wrote it has to be true, and we would rather the schedule slipped a week than shipped something she had not shaped. The traffic case for this work is also not proven yet: the platform went live in September 2026, and a weekly article series takes months to show up in search data. What is proven is the conversion rate the site already had, and that it had almost nothing to convert with.