TikTok creative: test phone-shot hooks first, or polished brand spots?

What order should you test TikTok creative in?

Teams often lead with a polished brand spot because it's the easiest thing to get signed off internally. But the feed punishes anything that looks like an ad in the first two seconds, so the polished version often loses on hook rate before the message even lands.

My view is to test creator-style, phone-shot hooks with on-screen text first, then build a higher-production version around whatever wins. The cheap versions give you a hook-rate baseline to judge the expensive ones against, and you learn faster.

The counter-argument is fair, though. Native-looking content can pull in low-intent clicks, and if your product needs trust or a price explanation, a polished spot may convert better even with a weaker hook rate. So it comes down to whether you judge creative on hook rate and CTR or on downstream conversions, and whether you have enough volume to tell the difference.

Which metric do you trust for the first round of tests?

Sasha, I'm not sure hook rate should be the first-round metric at all. It only tells you someone stopped scrolling, not whether the right people are arriving. A phone-shot hook can win on 3-second views and then send visitors who bounce off the landing page, which shows up as cheap clicks and a poor cost per conversion.

Before launch, write down your primary metric and a minimum conversion count per variant. If one variant gets 20 conversions and the other gets 4, hook rate is only a tiebreaker. Also compare landing page engagement in GA4 against platform CTR, split by creative variant through UTM parameters. If the creator-style hook has a similar conversion rate and a similar engagement rate, that's a real win. If its sessions engage far less, you've found the low-intent problem you raised.

Decay matters too. Novelty formats often lose hook rate after a couple of weeks as the audience gets used to them, so a round-one winner needs a re-test before you build the expensive version around it.

GA4 counts a session as engaged if it lasts 10 seconds or more, has two or more page views, or includes a conversion. So someone who lands from a phone-shot hook, skims for six seconds and leaves counts as unengaged, even if they were exactly the buyer you wanted and would have converted on a second visit. That means part of the engagement gap Kai describes is just the definition at work. I'd treat engagement as a diagnostic and let conversion rate per session pick the winner, with the landing page identical across variants so you aren't testing two pages at once.

For the UTMs, TikTok's URL macros for campaign, ad group and ad IDs are cleaner than hand-typed tags per variant, since nobody can introduce a typo when copying a link.

I'd push back on the two-week decay rule, because calendar time is a poor proxy for fatigue. What wears out a hook is how often the same people see it, so frequency is the number to watch. Say a dental clinic runs a phone-shot hook to a small local audience at high daily spend. The same faces can see it five or six times in a week, and hook rate can slide within days. That same hook run to a broad audience at modest spend might hold for a month. A fixed two-week re-test would retire some winners too early and keep dying ones live too long.

I'd set a trigger instead. Track hook rate against frequency for each variant, and re-test when hook rate drops 20 to 25 percent below its launch level at a comparable frequency. That gives you something concrete to act on.

Agreed on the frequency trigger, but there's one thing that can quietly break it. If both hooks sit in the same ad group, delivery tends to favor whichever one does well early, so one variant ends up with much higher frequency and you never get a comparison at matched frequency. Put each variant in its own ad group with a set daily budget so they reach the same audience at a similar rate. Otherwise you're measuring which hook the algorithm picked, not which hook actually works.

Sasha and Kai both say 3-second views, but TikTok's standard hook metric is the 2-second video view, with 6-second views as the other common one. Meta counts at 3 seconds, so if you pull numbers from both platforms into one sheet, the TikTok hook rates won't be comparable. Pick one definition, copy the exact metric name from the export into your column header, and use that same field for the frequency trigger. A 20 to 25 percent drop only means something if the launch baseline and the current value come from the same field.

Also, once the TikTok URL macros are set up, fire a test click and open the landing page URL to confirm the parameters actually filled in. An unresolved macro can show up in GA4 as a literal string or as (not set), and then your variant split is quietly wrong.

Before anyone argues about which metric to use, set one rule: swap only the first few seconds and keep everything after the hook identical. If the phone-shot version also has a different middle, offer, or call to action, a change in conversions could come from any of those, and you can't tell which. Cut both hooks from the same master edit and attach the same body and ending. Then any gap you see comes from the opening, which is the thing you're trying to learn.

Sample size is usually where this stalls. Twenty conversions per variant sounds reasonable until you run the numbers on a modest budget. If each lead costs around 40 dollars and both hooks get half the spend, you could be waiting several weeks. In that case, use a higher-volume step like a form start or add-to-cart as a provisional signal, and only trust the final conversion count once it's in. Otherwise you'll either call it too early or never call it. Roughly what does your cost per conversion look like?