What the founder believed walking in
A clinician runs a mobile, cash-pay house-call practice. Practitioners go to people’s homes when they are sick, in a dense city. After the part of the business that used to pay the bills dried up, she needed local patients, and she needed them soon.
So she did the reasonable thing. She looked for a marketing problem, because that is what it felt like.
Her fear was about the budget. Is a Google-led bet worth the money, when a marketer had already looked at her and said there was no demand and he would not take her money? How would she even know if the spend was working? She had an account she had inherited, and it reported a 2% conversion rate. On paper the ads were doing fine. She could not understand why the phone stayed quiet.
None of that is a foolish way to think. A 2% conversion rate is a normal, healthy-looking number. If your one visible metric says the marketing works, the natural next question is how much more to spend. She was reading the only scoreboard she had.
What we saw
We looked at the scoreboard before we looked at anything else. It did not survive the look.
That 2% was not a real 2%. The account was counting things that were not appointments. A phone-number click counted. A page refresh counted. We showed it live, on her own screen, by refreshing the page and watching the appointment count tick up by two. A phone call is a real signal, but it is not a booked visit. It is a micro-conversion, not a true one.
The number had been set up by someone who did not know better. That is the good news, because the only direction from there is up. But sitting on a fake 2% is worse than sitting on an ugly true number, because a fake number tells you to keep spending into a machine nobody has actually measured. You cannot fix a leak you have been told does not exist.
Here is the part that matters. She did not have a demand problem, and she did not have an ads problem. She had a measurement problem sitting on top of a conversion problem, and both were invisible because the account was lying to her in a language she had no reason to distrust.
What we decided, and what we chose not to do
The honest thing to say is that we did not know the real number yet. When she asked what to expect, the answer was that we would establish a baseline first, not a prediction. We would rather tell you 150 and come in at 90 than tell you 90 and come in at 150.
So the first job was not growth. It was truth. Rebuild the measurement so every number after it could be trusted, then fix the path the buyer actually walks, then decide anything about spend.
And we chose not to touch her in-progress internal systems. The practice was mid-build on its own operational tooling, and it would have been easy to fold that into the project and bill for it. We left it alone. The pilot was about getting her one step closer to being seen by the right patients, not about building things she had not asked for. Spend was treated as a capped input to draw down only as real conversions justified it, not a quota to burn.
What we built
We rebuilt the plumbing first.
- True-conversion tracking. Redefined a conversion as a booked, paid visit, not a click or a refresh. Consolidated the analytics into one place so there was a single source of truth.
- HIPAA-compliant call tracking, so a real phone booking was counted as one, and a click on the number was not.
- Account hygiene. A negative-keyword list, blacklisting for click fraud, and excluding the practice’s own staff so internal traffic stopped inflating the count. We also moved the site and the ad accounts into the client’s own ownership, so every asset was hers to keep.
Then, with an honest baseline in hand, we rebuilt what the honest number exposed.
- Price-forward ads that disqualify. We put the price into the ad itself, so people who were never going to pay filtered themselves out before they cost a click. The exclusions went in too.
- A conversion-engineered landing page to replace the inherited page, which had been converting at about 1%. Price up front. Proof and hours above the fold. Plain, non-clinical language for the symptoms. A low-friction callback instead of a payment wall, and a purpose-built mobile layout instead of a desktop page squeezed onto a phone.
What changed
With real tracking underneath it, the work started to move real numbers.
Putting the price in the ad roughly doubled click-through, from around 2.x to around 4.6. That is fewer wasted clicks and more of the right ones. The rebuilt page and ad structure lifted the conversion rate from about 1.95% to 6.25%. Not a fake 2% that meant nothing, but a real 6.25% that meant booked visits.
Across the full run the practice booked 247 appointments, about 10 over the 237 the team had forecast, and about $80K in collected first-visit revenue. These are the results from one specific engagement, measured by us, not a third party, and not a promise of what your account will do.
What a patient really cost
The dishonest scoreboard had one more number in it. On the dashboard, an appointment looked like it cost roughly $11 to win, and an $11 patient is the kind of bargain nobody thinks to question. That $11 was fiction for the same reason the 2% was, because the account was counting refreshes and phone taps, not patients.
Honest tracking replaced the $11 with a number nobody enjoys seeing: roughly $250 for every patient who actually booked. That’s the trade honest measurement makes. The comfortable fake goes away, and what comes back is an ugly true number you can finally work on. Once the tracking, the ads, and the page were rebuilt, the true cost came down to roughly $76 per booked patient, about a 70% decrease. The drop is measured from the honest $250, not from the $11, because the $11 was never a cost at all.
The deeper economics of this practice, whether a cash-pay model could carry the business over time, and the separate work of bringing dormant patients back, are their own story and not this one. What we can say is that the practice went on to scale, and later expanded into telehealth.
The lesson
When a founder says the marketing is not working, the marketing is usually not the first thing to look at. The first thing to look at is whether the numbers are even real, because most performance numbers handed over by a previous vendor are not.
A flattering metric is more dangerous than an ugly one. An ugly number tells you where the leak is. A pretty fake one tells you to pour more money into a machine nobody has measured, and that is the most expensive kind of comfortable.
Data does not lie. That is the part worth trusting. The numbers may not be the ones you were hoping for, but they are consistent, and you can build on them. You cannot build on a number that was invented to make you feel fine.
So before you decide the ads are broken, or the market is dead, or you need to spend more, there is one quieter question worth asking first. What would your own scoreboard say