Hook rate on Meta video ads: rank creatives by it, or only use it to diagnose?

I'm not sure 3-second plays and hook rate deserve as much attention as they get when reviewing new video creatives. Take a dental clinic running two videos. One opens with the logo and a smiling team shot, the other with a patient saying the price out loud. The first gets the better hook rate, but the second gets more link clicks and more booked calls. I'd trust the second every time.

Hook rate only tells you the opening seconds held attention, not that anyone wanted what came after. So rank creatives on cost per result, and use hook rate as a diagnostic for figuring out why one is losing. Do others rank by hook rate first, or only look at it once something underperforms?

Ranking by cost per result makes sense. Hook rate gets more useful once you hold everything else constant and swap only the first three seconds. Then the hook is the only variable, so a higher hook rate really does mean that opening stopped more thumbs. Comparing two different videos on hook rate is where it misleads. A logo-and-team opener gets played through by people who were never going to book, which inflates the number.

I'd run the price-first opener against the logo opener with the same script after it, judge on cost per booked call, and use hook rate to explain why one version pulled fewer but better plays. At low spend, CPA swings so much that the hook test can tell you more early on.

How many conversions does each creative need before you trust a CPA gap?

Thirty conversions per creative is the usual rule of thumb, and it's a fair floor. But the size of the gap matters as much as the count. At 10 or 15 booked calls a side, a 25% CPA difference is often just noise, so I'd hold off calling a winner. A new opener also changes who the algorithm shows the ad to, so even with the same script after it, the two versions aren't a clean test. Putting them in separate ad sets with equal budgets keeps delivery differences from muddying the hook comparison.

Separate ad sets is the right call. I'd also launch both on the same day and leave both budgets alone until they've exited learning. Meta resets learning when you edit an ad set, and a version that catches a cheap auction early can look like a winner for days.

A good rule of thumb is to set the stop condition before launch, something like 30 booked calls per side and at least two full weeks. Only call a winner if the CPA gap holds across that whole window. If it flips in week two, you were looking at noise.

Depends on how much volume you can actually push through each side. If the creative is pulling a few hundred plays a day across both ad sets, two weeks gets you to 30 calls fast. For a local clinic on a small budget it could take six or eight weeks, and by then seasonality and the booking calendar will matter more than the test itself.

Kai's right that volume decides it, and what matters most is how many booked calls per week the account can produce per side. Once you know that, the stop rule is just arithmetic: 30 calls divided by weekly calls per side gives you how long to run the test. If that comes out to more than a month, a hook test on a small clinic probably isn't worth the wait.

The math checks out, but capacity matters as much as bookings. A small clinic can fill its chairs partway through a six-week test, and once that happens the call count stops telling you anything clean about the ad. You'd be measuring how full the calendar is, not how well the ad works. Look at open calendar capacity first, then decide if the test is worth running.

Agreed, and the lag makes the calendar problem worse. Dental bookings often land days after the click, so week one's count keeps growing into week two, and a full calendar quietly drops the late ones. I'd judge the test on the enquiry instead. Count the calls or form fills that reached the front desk on each side, and log whether each was booked, waitlisted or turned away. The enquiry count shows what the creative earned. The calendar only decides how many of those become patients.

Logging the enquiry is the right move. The one thing I'd add is a check on booking rate, because a price-first opener tends to pull people who just want a number. Its enquiries can look cheap while fewer of them actually book. Work out cost per booked patient for each side: cost per enquiry divided by the share of enquiries that book. If the price hook brings enquiries in 20% cheaper but only half of them book, it's the more expensive version, even though the front-desk count makes it look like the winner.

Depends on the enquiry-to-booking lag. The booked share is the number that matters, but as Iris pointed out, bookings land days after the click, and a patient shopping on price can take a couple of weeks to decide. If you read the booking rate on day seven, the price-first side can look worse just because its people are still deciding. Set a cutoff, say 14 days after the enquiry, and compare both sides only on enquiries old enough to have finished.