Meta 1-day view attribution: count it for budget decisions or only click-based results?

Picture a dental clinic running Meta lead ads at a steady pace. Default reporting looks fine, but filter to 7-day click only and cost per lead jumps around 40 percent. That gap is view-through: someone watched a video, didn't click, then booked a few days later, probably after searching the clinic name. Some people argue view conversions are real influence and should count, especially for higher-ticket services. Others think they flatter the platform, since Meta takes credit for bookings that would have happened anyway. My take is that the number isn't wrong, it's just not one to bid on. Use click-based numbers for budget decisions and treat view-through as a sanity check that the creative is doing something. Which window are you all using for the weekly review?

I'd drop the 1-day view number from budget decisions entirely. It mostly catches people who were already leaning toward booking and just happened to scroll past an ad that day, so the platform is basically grading its own homework. The 7-day click number is stricter, but it still can't tell you whether the ads caused the booking. The only view-through I'd trust comes from a holdout or geo split, where part of the audience sees no ads and you check whether bookings actually drop. Has anyone run a holdout on a lead campaign and seen how much of that view-through was real?

Agreed on the holdout. One thing about the test itself: a geo split on a lead campaign often falls apart because lead volume per region is small, so a week of bookings can swing on noise alone. It usually works better to run it longer than feels reasonable, or to match regions on past lead volume before you split them. And if the holdout does show a drop, look at whether it's leads or booked appointments. They can move differently, and lead count is the one that gets inflated by view-through.

Noise is a real issue, and for thin regional volume I'd try a different tool first. Meta's Conversion Lift study (in Experiments) randomizes at the user level, so you don't need two regions with matched lead counts to get a readable result. It still needs enough conversions and a budget you can hold for a couple of weeks, so a very small clinic might not clear that bar either.

The lead versus appointment distinction is where I'd spend more time. If the lead form is your conversion, the lead lives inside Meta, and that's exactly the number view-through inflates. Pull the booked appointments from your CRM and match them back to the leads, even roughly, before you judge the holdout. Otherwise you're grading the test on the same number it's supposed to check.

Conversion Lift doesn't get around the lead problem. It measures whatever conversion event you point it at, so if that's the lead form, the holdout is still grading leads. To fix it, you'd need booked appointments going back to Meta as an offline or CAPI event first. Otherwise the lift number has the same blind spot.