Does removing health details hurt Facebook ad optimisation? What we know and what we don't

The most common objection I hear to cleaning health data is simple: "won't the ads perform worse if I send less?"

It's a fair question, and it deserves a straight answer rather than reassurance. So let me split it into what I know, what I don't, and how I'd find out for your account.

What I know. Matching relies mostly on who the person is and which ad they clicked. If you keep sending a hashed email, a hashed phone and the click ID, the platforms can still match the conversion to the right person. Removing the page path and the descriptive event name doesn't remove those. So the mechanism that ties a conversion to a person survives the cleaning.

What I don't know. I don't have a controlled study across many accounts showing how optimisation changes when you neutralise event names. Delivery systems are complicated, and I'd rather say that plainly than dress up a guess as a fact. If someone gives you a precise percentage, ask where it came from.

What changes for sure: the alternative. The comparison people make is "full data and great performance" versus "clean data and worse performance". For a lot of clinics, that isn't the real comparison. The real one is "descriptive data that gets held back" versus "clean data that arrives". When events are blocked or stripped, match quality drops, cost per lead climbs and campaigns stall. Sending less can be better than sending something that gets dropped. You're comparing clean, complete signals against partial, unreliable ones.

So how would I find out for a specific account? A simple experiment.

One: pick one campaign, not all of them. Changing everything at once means you can't tell what caused what.

Two: write down your metrics before you change anything. Cost per lead, cost per booked consultation, cost per attended consultation, and how many events Meta actually received.

Three: make the change, and don't touch budgets or creative for a few weeks.

Four: compare against the before numbers on the metric that matters. Not cost per lead, which can mislead, but the cost of the stage you care about.

Five: check the events themselves. Is Meta receiving more of them, and is match quality steady or better?

If the numbers get worse, you have data, and you can roll back. If they get better or stay steady, you've gained safety without a cost. Either way you've replaced an argument with a result.

There's also a second-order benefit that's easy to overlook. Once the data is clean, you have room to send better signals, like booked and attended stages, which teach the campaign more than the form ever did. So the cleaning isn't only about avoiding a problem. It can be the first step toward a signal that optimises better.

A few notes on making the experiment fair. Run it long enough to see the effect: for most clinics that means at least a few weeks, and longer for high-consideration services where the cycle is slow. Keep seasonality in mind: comparing a slow month to a busy one tells you nothing. Don't change the budget, the creative or the audience during the test. Track the same metric before and after, and write your expectation down before you look at the numbers.

And decide in advance what would make you roll back. If cost per attended consultation rises well beyond what you'd tolerate, or if Meta receives fewer events, you'll want a clear line, not a debate after the fact.

If you've tested this: what happened to cost per booked consultation? Not cost per lead. The stage that actually matters.