Facebook ads for dentists: why the leads are cheap and the consults never seat

The Meta numbers in most full-arch accounts look fine, which is the problem. Cost per lead is low and volume is climbing. Then you ask the front desk what happened to last week’s leads, and the answer is some version of: most never picked up, and the ones who did were not candidates.

Nothing is broken. The account is doing precisely what it was configured to do.

The account optimizes toward the event you report

Meta’s delivery system is not looking for people who want implants. It is looking for people who resemble the ones who already triggered your conversion event. If that event fires the moment a form is submitted, the model spends the next three weeks getting very good at finding people who submit forms. Whether any of them sat in a chair is information it never receives.

This is the most consequential setting in the account, and in most practices it has never been changed.

The version that works reports a later event: consult booked, or better, consult seated. Which means the booking has to exist somewhere you can export it from, and that makes this a records problem before it is an ads problem. If a seated consult only exists as a block in the appointment book, there is nothing to send back.

A practice that cannot wire this up automatically can still upload offline conversions by hand. A weekly file with the lead’s email, the event name and a timestamp will move delivery. It is tedious, and it does more than any audience change you could make instead.

Instant forms are frictionless by design, and that is the leak

An instant form prefills name, email and phone from the person’s profile. Two taps and no typing, without ever leaving the feed. Plenty of people who submit one do not fully register that they have contacted a dental practice at all.

Friction is not the enemy here. Friction is the filter. Every step you remove between the ad and the submission removes information about how much the person actually wants this.

So put steps back deliberately.

Ask a screening question with real answers on it. “Are you missing teeth on the upper, the lower, or both?” and “Are you looking to start treatment in the next three months, or researching for later?” A candidate answers those without thinking. Someone killing ten minutes on their phone drops out.

State the cost range inside the form. Full-arch treatment lands somewhere in the $20,000 to $45,000 band depending on scope, and a form that says so will cut submissions sharply. The submissions it cuts are the ones your team was going to spend a week failing to reach.

Your cost per lead will go up. That is the intended outcome rather than a side effect, and it is why holding Meta to a search campaign’s cost per lead pushes an account in exactly the wrong direction.

Speed to lead has to be a route, not an intention

A Meta lead differs from a search lead in the one way that decides everything: the person was not looking for you. Twenty minutes later they have scrolled past a hundred other things and the moment has closed over.

Two things have to be true for this to work. The lead has to land somewhere a human is watching within minutes, not in a shared inbox that gets read after lunch. And one named person has to own it, with every attempt logged, because “the front desk will call them” is an intention rather than an assignment.

A workable standard looks like this. A text goes out automatically on submission. A human attempts a call inside five minutes during opening hours. The record closes with a reason after a defined number of attempts across the first two days, so nothing ages quietly in the middle. The conversation that qualifies the lead is a separate skill, and it never gets a chance to happen if nobody dials.

Most of your audience settings are doing less than you think

Practices spend their optimization energy here, and it is largely spent. Interest stacks built around “dentures” and “dental implants” mattered more several years ago than they do now. The delivery model reads response to your creative faster and more accurately than any layered interest list reads it for you.

Two settings still earn their place. Geography, sized to how far a full-arch patient will genuinely drive, which is further than your general dentistry radius and usually much further than a default ten-mile ring. And exclusions: current patients, and anyone already in your pipeline including past consults. Paying to re-reach the patient your own records list as “financing pending” is not remarketing.

The creative is the targeting. A video of a real patient describing what they stopped being able to eat selects its own audience more precisely than any interest setting will.

Read it as a cohort, not as a month

Meta’s lag and your month boundary do not line up. Leads generated in the last ten days of a month are still inside their decision window when the month closes, so every month-end report understates the most recent spend. Read enough of them and you will conclude the channel is degrading when it is only young.

Group leads by the week they first made contact, and follow that group forward through contacted, qualified, booked, seated, started. A cohort from six weeks ago can be read. Last week’s cannot, and belongs in the report labelled as incomplete rather than averaged in with the rest.

What falls out is a cost per seated consult you can actually defend. Say a practice puts $6,000 into a week of Meta and that cohort eventually produces four seated consults. That is $1,500 a seated consult, against a case in the $20,000 to $45,000 band and whatever the practice’s real close rate turns out to be. Now the argument about whether Meta is expensive has numbers in it instead of feelings.

The cohort view also exposes the ghost. Some share of the people your ads reached will never click anything, and will arrive weeks later as a direct visit or a branded search. Paid media measured only on click-attributed consults will always undercount social, and the gap widens the more of the funnel Meta is carrying. Asking every booked patient where they first heard of you, and storing that answer separately from the tracked source, is the cheapest correction available to you.