6/ When hold rate starts decaying (it will), the loop restarts at step 1. The best creative engine isn't "always test new." It's "squeeze every winner until it dies, then ship one new hypothesis."
I started studying the biggest US supplement accounts built around AI avatars.
Almost all of them use the same type of character: the herbalist grandma, the Amish farmer, the farmer with 50 years of experience, the monk. White-haired people who instantly inspire trust just by showing up.
Several of these accounts have over 400K followers, and most of them sell just one product through the link in their bio.
VOC sourcing, ranked by signal density for paid-social hooks:
1. Reddit threads (in your subcategory)
→ unfiltered buyer language, real objections
2. Amazon reviews (1-3 stars)
→ use-case edge cases nobody else writes
3. TikTok comments under competitor ads
→ emotional reactions in real time
4. Meta ad comments
→ useful but already moderated by algorithm
Skip "review mining" of 5-star reviews. That's confirmation bias, not research.
Mine the 1- and 2-star. The complaint IS the hook.
@georgi2103 DTC Creative Strategist since 2020 in supplements and health. Strong on avatars, angles, hooks, and Meta creative tests.
portfolio: https://t.co/FbVtrG2SVp
Hook rate and hold rate are not the same metric.
Hook rate = did your first 3 seconds stop the scroll?
Hold rate = did your body earn the next 30?
Most CSs only track one. Then ship "winning" ads that crash at 1.5x ROAS because they never held.
If you're not reading both you don't know if your hook is winning or just borrowing time.