Dr.Reju-All generated 514 angles across 5 SKUs. We deleted 127 of them.
514 candidate angles generated across 5 K-beauty SKUs, 127 deleted, 387 kept with at least $5K of GMV behind each. How a culled angle library moved an 842-creator roster.
Source: app.sfn.ai / Reporting · Dr.Reju-All US · Apr 1 to Jun 30, 2026 · actual reported GMV
THE PROBLEMFive SKUs, 842 creators, and a category where everyone sounds identical
Dr.Reju-All runs a five-SKU Korean skincare line: a PDRN moisturizer, a retinol and niacinamide serum, a ceramide barrier cream, a lip serum and a sheet mask.
K-beauty is the most crowded creative category on TikTok Shop. Every competitor is saying glass skin, barrier repair, PDRN, collagen. A creator handed a product and told to be authentic will reproduce the category average, because the category average is all they have seen. Across 842 creators, the category average is what you get 842 times.
There is a compliance edge underneath it too. The serum is a brightening product, not a whitening one. That distinction is not cosmetic, it is the difference between a compliant claim and one that takes the listing down.
THE STRATEGYGenerate wide, then cut hard on evidence
We mapped 25 competitors, five against each SKU, chosen on real GMV rather than on who looked similar. Against the moisturizer alone that cohort represented roughly $23.2M in lifetime GMV. Against the serum, roughly $12.9M.
Then the engine read their creative and generated 514 candidate angles off that coverage, each one a pattern with a hook, a story arc, a funnel position and a commercial approach attached.
514 is too many. It is also the wrong number to optimise for, because a library you cannot trust is worse than a small one you can. So we culled on a single hard rule: every angle had to be sitting on at least $5,000 of average GMV in the source videos it was drawn from. 127 angles failed and were deleted outright. 387 survived, the weakest of them carrying $5,800 and the strongest carrying $290,807.
Anyone can generate a thousand content ideas. The discipline is throwing away the ones with no money behind them. A creator opening the app should never be looking at an idea that has not already earned somewhere else, because the entire premise of a daily idea is that it is a bet someone else already won.
The fence went up alongside it. The brand-level forbidden list ran to its full 100 entries: 88 policy and category terms plus 12 competitor brand names, so the roster could not drift into naming the exact K-beauty brands whose creative we had just studied. Each SKU then carried its own banned list for the foreign ingredients and formats its competitors have and it does not. The moisturizer blocks beef tallow, manuka and shea. The serum blocks micro-infusion. The lip serum blocks lip tint and peel-off masks. The brand's own actives were never blocked.
THE RESULTTwo flat months, then a curve that does not stop
April is nothing. Most of May is nothing. The first faint movement comes around the middle of May and it is measured in tens of dollars.
Then June. $499 and $430 in the first week, $615, $545, $806, $822, and it keeps compounding: $1.9K on Jun 17, $3.9K the next day, $4.8K on Jun 21, $6.0K, and a peak of $6.8K on Jun 24. The last week of the window never drops out of the $4.8K to $6.4K band, closing Jun 30 at $4.9K.
$74.0K in reported GMV across the three months, 11.2K orders, 2.5M views, 1.8K videos from 842 creators. Average order value lands near $6.60, which tells you this was a volume-and-reach program rather than a high-ticket one, and the strategy was built accordingly.
The two flat months are the part worth sitting with. That is what building a library looks like from the outside: nothing, nothing, nothing, and then a curve that does not stop climbing, because the roster finally has something specific to say.
Figures cover the Apr 1 to Jun 30, 2026 engagement window.
