Nobody knew what a patient cost
A multi-specialty clinic in Moscow: twenty departments, its own diagnostics, fifteen years of work. Five contractors ran the marketing and each one sent its own report. Not one of them could answer what a single patient cost.
Campaigns, SEO, marketplace listings and the site migration were handled by the team.
What was wrong
Search advertising lived in one spreadsheet, SEO in another, the booking marketplaces in a third. Impressions, clicks, cost per click - each had their own, and each was right in their own way.
The reports agreed on nothing except that they were all about clicks. The question none of them answered: what does a patient cost. Not an enquiry, not a call, but a person who reached the appointment and paid.
Until that answer exists, every budget decision is made by feel: add here, cut there - and what comes of it will be clear later and imprecisely.
What I did
For the first few weeks we left the advertising alone. First we had to see the whole path a patient takes: click, call, booking, visit, payment. And to join the ends - for every call to know the ad it came from, for every payment to know the call that started it. That is what end-to-end analytics and call tracking do.
The first thing that became visible: a third of the campaigns produced cheap clicks and not one patient. Those were switched off. Yandex.Direct was rebuilt so the algorithm learned from bookings rather than from visits: within the first month the cost of an enquiry fell from 1,494 to 986 rubles.
Then we dealt separately with something that does not look like marketing at all. Half the calls never became a booking - a person would call, ask a question and hang up. That is not advertising, that is how the front desk talks, and it gets fixed by listening to recordings and working with people. Seven callers in ten started booking, instead of four.
Search and the marketplaces where a patient chooses a doctor came last: maps, 2GIS, ProDoctorov and the rest. Over eighteen months free search traffic grew from 1,200 to 4,620 visits a month - patients you do not pay for again every time.
The system was not built for the sake of a report. When you can see the cost of a patient by channel, budget moves at the moment the numbers shift - not at the quarterly review, when the money is already spent.
I treated the migration to a new platform as a technical job: move the structure, set the redirects, hold the rankings. The redirects went in and traffic still dropped, and it took a couple of months to come back.
With hindsight the thing to budget was not a week of checking but a quarter of recovery - and to warn the clinic about it up front instead of explaining the dip after the fact. The migration itself was needed: the old site loaded four times slower than the new one.
Five contractors is not a story about bad contractors. Each of them did their own job honestly and showed real numbers. None of them owned the whole, and the question of what a patient costs lives exactly there.