An ad platform learns from whatever you tell it counts. Tell it that every form fill counts, and it gets very good at finding people who fill in forms. That's not the same as finding people who start a full-arch case.

This is one of the most common reasons full-arch campaigns drift. The leads keep coming, the cost per lead looks fine, and the consult room fills with people who were never close to a decision. Nothing is broken. The campaign is doing exactly what it was told.

What the ad platform is actually optimising for

Meta and Google don't know what a full-arch patient is. They know what you report back to them as a conversion, and they go looking for more people like that.

If the conversion you report is "submitted a form", the platform studies everyone who submitted a form and finds more people who behave like them. Some will be serious candidates. Many will be price shoppers, people outside your area, people who tap every health ad they see, or people curious about the cost who were never going to book.

The platform can't tell those groups apart unless you tell it which ones turned into patients. So it optimises for the thing it can see.

Why "more leads" made it worse

This is where the reporting makes it worse.

When the agency's report leads with cost per lead, the natural move is to push that number down. The cheapest way to get cheaper leads is to let the platform find people who fill in forms easily. Those are rarely the people who go on to accept a large treatment plan.

So the report improves every month. More leads, lower cost per lead. Meanwhile the front desk is calling people who don't pick up, the consults that do get booked don't show, and the surgical day still has open time on it. If a metric can go up while your bank account doesn't, it doesn't belong in a promise.

Agencies paid on leads have no reason to fix this. The report looks good either way.

What to send back instead

The fix is to change what the platform learns from. Instead of reporting every form fill as the win, report the stages that actually matter to the practice: the consult that was attended, and the case that started.

That takes tracking that actually works, end to end:

  • Analytics and a tag manager set up so every inquiry is recorded once, with the page and campaign it came from.
  • Call tracking, because a lot of full-arch patients call rather than fill in a form, and a call longer than thirty seconds counts as an inquiry.
  • Campaign tags on every link, so each inquiry can be traced to the ad that produced it.
  • Server-side conversion events, so the platform hears about the stages further down, not just the first click.

Once the platform starts hearing about attended consults and started cases, it starts looking for people who resemble those patients instead. You can't do that if you're paid on leads, because it makes your lead numbers look worse before it makes the practice's numbers better.

This is part of how I run Meta ads for practices, and it's why the stages after the lead are counted separately on /how-we-measure.

The HIPAA question

The obvious worry is whether any of this sends patient information to an ad platform. It shouldn't, and in what I set up it doesn't.

No patient record goes to Meta. What goes back is a hashed, server-side conversion signal, not a patient record, and nothing from your EHR is uploaded. Matching against the practice's records happens on the practice's side, with the minimum information needed and under a business associate agreement where patient health information is involved. I explain the setup on how I handle HIPAA in marketing. If you have a compliance person, I'd rather they were on the call.

Dashboards always look fine

The other reason this goes unnoticed is that a broken or misdirected campaign rarely looks broken.

On one client account I found a conversion tag that had been feeding false data into every optimisation decision for months. Nobody caught it, because the dashboard looked fine. The numbers went up and to the right. They just weren't measuring anything real.

That's the risk with any campaign judged on the top of the funnel. The dashboard can look healthy for a long time while the platform is being taught the wrong lesson.

What this can't fix

Sending better signals back changes who the ads find. It doesn't change what happens after that.

If inquiries aren't answered quickly, better leads still go cold. If the consult isn't converting, better leads still leave thinking about it. Those stages are covered in from consult to started case, and they matter as much as the ads.

It also doesn't make a weak case economically viable. If a started case isn't worth comfortably more than it costs to acquire, which you can work out in what a full-arch case costs to acquire, no amount of better targeting fixes the arithmetic.

That's why the full-arch program is judged on one number: fifteen paying full-arch patients in ninety days. If you miss it, you get the entire fee back and I keep working at no fee until you have fifteen, while the agreed ad budget is funded.

Full-arch ad questions

Is it HIPAA-compliant to send conversions back to Meta?

It can be, if it's set up properly. No patient record goes to Meta. What goes back is a hashed, server-side conversion signal, and nothing from the practice's EHR is uploaded. Matching against practice records happens on the practice's side under a business associate agreement where health information is involved.

How long before the ads start finding better patients?

It depends on how many conversions the platform hears about, so a practice with more consults gives it more to learn from. I don't put a day count on it. Judge the campaign at ninety days on attended consults and started cases, not on the first two weeks.

Should I stop paying for leads?

Stop judging the campaign on leads. Leads are still the first step, but cost per lead counts the cheapest thing in your funnel. Judge it on attended consults and started cases, and make sure the platform is learning from those too.

Why are my full-arch leads so hard to reach?

Often because the campaign was taught to find people who fill in forms easily, and many of them were never close to a decision. Slow follow-up makes it worse. Both are fixable, and they're separate problems.