TikTok ads: swap only the hook on a flat video, or ship a whole new cut?

I keep running into the same question with TikTok creative tests. When a video underperforms in the first few days, do you swap just the opening 2 seconds and keep the rest, or build a whole new video with a different angle?

Take a supplement brand with a strong middle but a flat hook. Swapping the hook is cheap and isolates the part that matters, because most drop-off on TikTok happens before second three. But if the body is what's weak, a new hook just buys a few more seconds of people watching a slow pitch.

My view is to do hook-only swaps first. They give a cleaner read on scroll-stopping and cost almost nothing to produce, so I'd only rebuild the whole video once hook tests stop moving watch-through. Does anyone see it differently at low spend?

Your order makes sense, but I'd let the retention curve decide what to fix instead of running hook tests by default. If most viewers are gone by second three, the hook is the problem. If the curve holds through the opening and then slides at a specific point, the body is the weak part, and a new hook just pushes that slide later. At low spend you rarely get enough impressions to separate two hooks cleanly anyway, so when the drop is in the middle, put the budget into a new cut.

Retention-curve test makes sense to me. I'd also compare the 2-second view rate with the 6-second one before deciding anything. Say a supplement video holds 2-second views fine but loses most viewers between seconds two and six. That means the hook is working and the start of the body is what's losing people. Swapping the hook won't fix that, and you'd only learn it after a week of spend on a test that couldn't have worked.

Holding 2-second views doesn't make the hook fine. A hook can stop the scroll and still promise something the body never delivers, so people bounce the moment the pitch shifts. Before you decide which part to swap, look at what's actually said in seconds two to six.

Mira's right about the promise gap, and you can check it quickly: write down what the hook claims, then write down the first thing the body actually says, and see if they line up. A decent rule of thumb is that if 2-second views look healthy but 6-second views drop off, fix the promise in the hook before you touch how the hook looks.

Honestly, I think the fix is even simpler than a new hook. When the hook promises a result and the body opens with a mechanism or an ingredient list, people drop off while they're still waiting for the payoff. The cheapest edit is usually to pull the first piece of proof (a customer quote, a before-and-after, the one number that backs the claim) up into seconds two to four, so the body answers the hook right away. Keep the hook's wording and visuals identical, because that way the test isolates the body change. Otherwise you're comparing two different ads again.