Polished brand video or phone-shot creator cut: which one do you test first on TikTok?

Picture a meal kit brand with one polished 30-second brand video that's run on TikTok for a month with flat CPA. The team wants to brief three more polished cuts. I'd test a phone-shot, creator-style version first instead, same offer and same audience. TikTok users scroll past anything that looks like a TV spot within the first second, so polish often reads as an ad and gets skipped.

That doesn't mean brand video never works. It can carry retargeting, or a launch where the name is already known. But for cold traffic, the first test budget is better spent on how the opening feels than on color grading. Has anyone seen the gap close, or does the polished version keep winning on your accounts?

Framing this as polished versus phone-shot is the wrong test. Those two cuts differ in pacing, captions, talent, lighting and length all at once, so if the phone version wins, you know the bundle won but not which part did the work. On a budget test that matters more than the style question. I'd hold the first three seconds and the offer line fixed and change only the opening, say a person talking to camera versus a close-up of the food. That tells you whether the hook is the lever before you spend anything on production.

Isolating the opening is right, but I'd settle the readout before the budget. CPA is a weak first signal for a meal kit, because one cell might only produce a handful of purchases in week one. The faster read is the 2-second and 6-second view rates (video_watched_2s and video_watched_6s divided by impressions), both in the TikTok ad report. If the person-to-camera opener keeps more people past second two, you have evidence the hook works before any orders land.

Also make sure both cells get comparable delivery. If the two creatives share one ad group, the algorithm can starve one early, and then you're comparing rates on very different sample sizes. Separate ad groups with matched budgets keep the comparison clean.

Jun, I don't think the 2-second view rate shows the hook works. It mostly tells you the first frame stopped thumbs, and a face or a bright close-up can do that with anyone, including people who will never order a meal kit. A talking-head opener can win on 6-second retention and still lose on link CTR, because viewers got curious about the person, not the offer. I'd use view rates as a gate, then pick the winner on link CTR and cost per purchase, with the threshold written down before launch.

Agreed on the gate, and I'd add one more rule before launch: nobody looks at cost per purchase until each cell has about 50 link clicks, and the CTR gap should hold in both halves of the run. Early clicks swing a lot, so a cell that's ahead after three days often isn't ahead after ten.

If the talking-head wins on retention but loses on link CTR, I wouldn't drop the style yet. Viewers are staying for the person, so the next variant should put the price or the offer on screen in the first three seconds. That's a second cell to test, not proof the format failed.

Putting the price on screen as the next cell changes a second thing on top of the opener. If the talking-head loses on CTR, you can't tell whether the face was the problem or the price overlay was what fixed something. It's the same bundling issue Kai raised, just one step later. Keep the person opener and the spoken offer line identical, and test only the first-three-seconds text in a separate cell, so each comparison has one variable.

Also, before you read CTR at all, make sure both cells count clicks the same way. TikTok's link clicks and GA4 landing page views often differ because of redirects and slow loads. If the UTM template isn't identical in both ad groups, the 50-click threshold means less than it looks.

Jun, I don't think the price overlay is really a separate variable here. In a 30-second-or-shorter TikTok opener the on-screen price is often the offer itself, so splitting "face" from "price text" ends up testing the wording, not the format. I'd run the text as its own cell against the same face and keep the spoken line identical, like you said.

Elena's right that the price on screen is the offer, so I'd stop debating how to label it and use a rule that holds either way: run two openers per round and let the winner carry into the next one. Keep the live cells small enough that each can reach about 50 link clicks within a week at your budget. If it can't, you have too many cells, not too little time.

The reason is that every extra cell splits the same spend. On a modest meal kit budget, a four-way test can sit under the click threshold for weeks while the creative starts to fatigue. Two cells finish fast, and a quick answer you can act on beats a clean one that shows up after you've moved on.

Two cells per round is the right call. I'd add one rule for judging it: pick the winner on the metric you wrote down before launch, and don't swap metrics because a different number happens to favor the cell you like. Say a talking-head wins on 6-second retention but loses on link CTR. Go with the CTR winner or rerun the round, rather than crowning whichever one looks nicer on the dashboard.