Built 2026-10-27 from the Smooche teardown (Meta Ad Library pull of 2026-09-07). Internal.
The scaling model, read off Smooche
Smooche (smooche.com, cosmetics) is the most complete of the five brands we took apart. It has the longest history (first product Feb 2024, peak Aug 2026), the most ads (7,162 linked to their domains), and it is the only one that used every lever we have names for. This page lays out what they did in order, week by week from May 2025, then turns it into a model with targets you can measure against.
~19 months
from the first product to the week the ad engine switched on (Feb 2024 → Sep 2025)
~11 months
from engine on to peak volume (Sep 2025 → Aug 2026); ads per week went from ~40 to ~430
1 upgrade
they launched the foundation on the pitch it already sold on (shade match), upgraded once to 50+ anti-aging, then held it
Read these three limits before the numbers.
- Ad counts are not spend. Everything here counts ads they launched. Meta does not publish spend for these ads, and we have no revenue figure for Smooche. Volume tells you what they did, and only indirectly what worked.
- Meta deletes old retired ads. Ads that stopped before roughly mid-2025 are mostly gone from the library, so weeks before September 2025 undercount, and the early "days alive" numbers only show the long-lived survivors.
- Doors and products are read from the ad text and link. These are keyword matches (a floor, not a census). The page publish dates come from Smooche's sitemap and are month-level unless a day is given.
1 · The model on one screen
Every brand we took apart runs the same loop. What differs is which dial they turn first. Smooche turns the product dial first, because it sells many small cosmetics, then turns every other dial on the one product that took off.
0 · Borrow a proven pitchColour-changing foundation already sold on TikTok Shop as "adapts to your shade". Smooche launched on that, then upgraded once to "the formula ages you" for women 50+.
→
1 · Find the hero productSmall tests on many SKUs. Five lines over 17 months; the 5th (foundation) won.
→
2 · Page first, then adsFoundation SKU Jul 8 2025, shade quiz page Jul 9, ads after.
→
3 · Volume of ads15 → 80 → 160 → 260 a week. Kill most within ~1 week.
→
4 · Widen on the heroNew doors (shade → 50+ → Botox → GLP-1), new people (woman 52, makeup artist, K-beauty fan), new advertiser pages, new bridge pages.
→
5 · Raise order value$59→$39 single, BOGO, $119 mega pack, post-click offer pages, sale-calendar closers.
The eight dials (our checklist's names) and what Smooche set them to
| Dial | The question | Smooche | How often it changed |
|---|---|---|---|
| WHY (argument) | What does the buyer want, and what is blamed for her not having it? | "Every foundation you own is aging you"; "smart formula instead of needles"; "Korean women my age use the formula". Before that, the launch pitch: "adapts to your skin tone", no more "close enough" shades. | Changed once. Shade match led in Dec 2025; 50+ anti-aging took over in Q1 2026 (242 vs 110 ads), then the two ran about 50/50 (Q3 2026: ~940 each). |
| DOOR | Which need does the same product answer? | Shade match → aging 50+ → Botox alternative → crepey neck / jowls → GLP-1 face; sensitive skin and rosacea in scripts. | One new door every ~2 months on the hero, each with its own bridge page. |
| WHO | Whose life is the ad about? | Woman 52 / 58 / 63, makeup artist of 20 years, Korean-beauty follower, Gen X; NL / DE / PL shoppers. | Same listicle body, a new headline per person. |
| CREATIVE | How is the ad made and how does it open? | Video and image ads carry the headlines. Catalog ads (product photo, name and price pulled from the store feed, mostly retargeting) run underneath the whole time. Revenge-reveal stories (500–1,500 words), makeup-artist rankings, demos, apology-sale closers. | Weekly. This is the main volume lever. |
| VOICE | Who says it, under which page name? | Their own page plus media pages they run: Cosmetic Times (1,436 ads), Korean Beauty Tips (1,367), Aging Queens Magazine (456); persona pages such as James Miami MUA. | New advertiser page every few weeks; 4 pages carry 97% of ads. |
| CHANNEL | Which auction? | Meta is what we can see. A reverse-google page means Google traffic too. EN / DE / NL / PL storefronts. | Rarely. |
| PAGE | Where does the click land? | PDP (product page) direct, 8-reasons listicle, "Korean pharmacy" advertorial, top-5 comparisons, quiz, pre-lander, post-click offer pages. 109 bridge pages in total. | Bridge pages: 1–3 new a week, bursts of 13–14 in a month. The PDP layout never changed. |
| OFFER | What does she take home, for how much? | $39 single (from $59), 4×$10 instalments, BOGO, $119 mega pack, $9 blender bump, $4.95 shipping protection, 30-day "if it doesn't blend" guarantee. | Every sale event; structure stable. |
2 · The stages, in order
Six stages. The dates come from the ad data and the sitemap; the "ads/week" figures are averages over the stage.
3 · Week by week
Every bar is one week of newly launched ads linking to Smooche, coloured by product line. Hover a bar for the detail; the numbered markers are the events listed under the chart.
New ads per week, by product line
Where the clicks were sent (share of the week's ads)
The PDP is the buy page. Bridge pages = advertorials and listicles. The bridge-page share rises from ~15% to ~50% once the foundation takes over.
Ad format: video vs catalog ads (share of the week's ads)
The ad library labels these "DCO". 2,408 of Smooche's 2,577 DCO ads show only {{product.name}} placeholders, which makes them catalog ads (Advantage+ catalog, the old DPA). Only 169 were dynamic-creative ads with written text.
Kill speed: median days alive of the week's ads that were switched off
Weeks before September 2025 (shaded) only kept their long-lived ads, so the early values overstate. From the engine week on, the typical ad was switched off in about a week, and in the peak weeks in 3–8 days.
Advertiser pages that launched ads that week
Their own page plus the media and persona pages they run. 19 pages in total linked to smooche.com.
Open the full week-by-week table (64 weeks)
4 · How fast do brands scale
Three brands, same method of measurement. "Engine on" = the first month the ad count jumped by an order of magnitude and kills dropped to about a week.
| Brand | Product | First product / first ads | Engine on | Time to engine | Peak | Engine to peak | Peak volume | First lever widened |
|---|---|---|---|---|---|---|---|---|
| Smooche | cosmetics, many SKUs | Feb 2024 (lip plumper) | Sep 2025 | ~19 months | Aug 2026 | ~11 months | ~1,240 ads/month, 434 in the top week | product line, then door + person on the hero |
| Resilia | oregano / garlic softgels | Jan 2025 | Aug 2025 | ~7 months | Sep 2025, again Aug–Sep 2026 | ~1 month | ~1,800 ads/month | door (24 symptoms on one softgel) + media pages |
| Sofyre | pest-repellent pouch | Aug 2025 (store existed since Nov 2024) | Mar 2026 (branching) | ~7 months | Aug–Sep 2026 | ~5 months | ~70 ads/day | person / place (home, farm, RV, cabin), then pest |
Reading: the slow part is finding the product-and-argument pair that pays. That took 7 to 19 months in all three. Once it paid, volume went 10× within 1 to 11 months. All three killed most ads inside ~5–10 days and lost ~40% of the doors they tried.
5 · Your questions, answered from the data
Did they start by testing different philosophies, then lock one?
They tested products first and never ran a set of competing philosophies. The foundation launched on the pitch the product already sold on across TikTok Shop ("the viral skin-tone adapting foundation"). Within about three months the 50+ anti-aging argument took over, and it came from buyers they already had: their brow, lash and lip pages had been aimed at aging women since 2025. After that one upgrade the argument held. The product order was lip plumper (Feb 2024), lip stain (Aug 2024), brow and lash serums (Nov 2024), foundation (Jul 2025), peptide serum (Jul 2026). Each small line got 1–4 bridge pages and a few hundred ads. Inside the winner they then tested doors and people, while the argument stayed fixed. For a single-product offer, the equivalent first test is several doors on the same argument.
They tested products first and never ran a set of competing philosophies. The foundation launched on the pitch the product already sold on across TikTok Shop ("the viral skin-tone adapting foundation"). Within about three months the 50+ anti-aging argument took over, and it came from buyers they already had: their brow, lash and lip pages had been aimed at aging women since 2025. After that one upgrade the argument held. The product order was lip plumper (Feb 2024), lip stain (Aug 2024), brow and lash serums (Nov 2024), foundation (Jul 2025), peptide serum (Jul 2026). Each small line got 1–4 bridge pages and a few hundred ads. Inside the winner they then tested doors and people, while the argument stayed fixed. For a single-product offer, the equivalent first test is several doors on the same argument.
How did they find the argument in the first place (stage 0)?
They borrowed it. Colour-changing foundation was already sold by many sellers on TikTok Shop (KIMUSE, Vima, Vibeourplus and generic listings, one showing 6.2K sold at $49.95), and "it adapts to your skin tone" was that category's pitch. Smooche's first foundation ads in December 2025 used exactly that pitch. They had done the same a year earlier with a page named
They borrowed it. Colour-changing foundation was already sold by many sellers on TikTok Shop (KIMUSE, Vima, Vibeourplus and generic listings, one showing 6.2K sold at $49.95), and "it adapts to your skin tone" was that category's pitch. Smooche's first foundation ads in December 2025 used exactly that pitch. They had done the same a year earlier with a page named
viral-peel-off-lip-stain-tiktok. The upgrade to 50+ anti-aging came from their own buyers. We cannot prove they saw the foundation on TikTok first; the timing fits and the product was not theirs to invent.Do they add more doors, more philosophies?
Doors and people, constantly; the argument changed once. On the foundation: shade match (Dec 2025) → aging 50+ and Korean pharmacy (Feb 2026) → Botox alternative (Feb–Mar 2026) → crepey neck, jowls, GLP-1 face (Aug 2026). People: woman 52 / makeup artist / Korean-beauty fan, each as a new headline on the same 8-reasons body. Most of the ad text ran on about 10 headline templates.
Doors and people, constantly; the argument changed once. On the foundation: shade match (Dec 2025) → aging 50+ and Korean pharmacy (Feb 2026) → Botox alternative (Feb–Mar 2026) → crepey neck, jowls, GLP-1 face (Aug 2026). People: woman 52 / makeup artist / Korean-beauty fan, each as a new headline on the same 8-reasons body. Most of the ad text ran on about 10 headline templates.
Is the landing page something they rarely touch?
Half right. The buy page barely moved: the foundation PDP kept the same 16 sections from Dec 2025 to Sep 2026 and only swapped the headline, the lead line, the authority badge and the review count. The bridge pages changed constantly: 109 of them, 13 new in December 2025 alone, the "reverse" advertorial in 7 versions, "bestfoundation" in 6. On the buy page they ran split tests through a token in the URL (
Half right. The buy page barely moved: the foundation PDP kept the same 16 sections from Dec 2025 to Sep 2026 and only swapped the headline, the lead line, the authority badge and the review count. The bridge pages changed constantly: 109 of them, 13 new in December 2025 alone, the "reverse" advertorial in 7 versions, "bestfoundation" in 6. On the buy page they ran split tests through a token in the URL (
?trybe=), so the ads never had to change.Do they launch a ton of ads every day?
At the end, yes: ~40–60 a day (250–430 a week) in Jul–Aug 2026. At the start, no: ~15 a week through summer 2025, ~80 a week in the first engine months, ~160 a week in spring 2026. The volume grew with the business, in steps.
At the end, yes: ~40–60 a day (250–430 a week) in Jul–Aug 2026. At the start, no: ~15 a week through summer 2025, ~80 a week in the first engine months, ~160 a week in spring 2026. The volume grew with the business, in steps.
Creators? Other platforms?
Creator-style persona pages exist (James Miami MUA 68 ads, Maggie's Beauty Secrets 67, Originaljenny.ugc, Jennifer Parramore), but they are small. The volume ran through media-style pages Smooche controls (Cosmetic Times, Korean Beauty Tips, Aging Queens Magazine), and Cosmetic Times advertised a month before Smooche's own page. Meta is the channel we can measure; a Google variant of the main advertorial exists, and the store sells in EN, DE, NL and PL. We have no TikTok read on them.
Creator-style persona pages exist (James Miami MUA 68 ads, Maggie's Beauty Secrets 67, Originaljenny.ugc, Jennifer Parramore), but they are small. The volume ran through media-style pages Smooche controls (Cosmetic Times, Korean Beauty Tips, Aging Queens Magazine), and Cosmetic Times advertised a month before Smooche's own page. Meta is the channel we can measure; a Google variant of the main advertorial exists, and the store sells in EN, DE, NL and PL. We have no TikTok read on them.
6 · The lessons, as rules
- Launch on a pitch that already sells, upgrade it once, then hold it.Smooche started the foundation on the TikTok Shop category pitch and moved to 50+ anti-aging within a quarter. After that, Smooche, Resilia and Sofyre all kept one villain and varied doors, people and pages, which are cheaper to test and reuse what already works.
- Finding the paying pair is the slow part; plan for months.7–19 months to the engine in all three brands. Until then the job is cheap tests and fast kills, with ~15–80 ads a week.
- Publish the page before the ads.Foundation SKU Jul 8, shade quiz Jul 9, ads in August. Sofyre's doors got their page 2–6 weeks before the ad wave.
- Once one product pays, put almost everything on it.Foundation went from 2% of ads (Sep–Dec 2025) to 61% (Dec–Mar) to 88% (Apr–Aug). The other lines were left to run on their own.
- Ads are the volume lever, and most die within a week.Median ad switched off after ~7 days at scale and 3–8 days at peak. Volume replaces them.
- Run catalog ads underneath everything, and put the volume into video.Catalog ads (feed-driven, mostly retargeting) were about half of all ads early and about 20% at peak. Video rose from ~40% to ~75% of ads and carried the headlines.
- Keep the buy page fixed; churn the bridge pages.One PDP layout for nine months; seven versions of the main advertorial; split tests on the buy box through a URL token.
- One listicle body, a new headline per person.The 8-reasons body was identical for the woman of 52, the makeup artist and the K-beauty fan.
- Run several advertiser pages, including media-style ones.Four pages carry 97% of ads; the media page started before the brand page. More pages means more auction room and less risk from one page being shut.
- Each new door gets its own page, and about 40% of doors die.Thyroid, menopause lashes, postpartum lashes, sensitive skin, Cleopatra never got a second version. Kill the ads, keep the page, try the next door.
- The story lane sells visible results.The top ad by engagement (92,732) is a revenge-reveal story: "At 63, I ran into my ex-husband at Walmart". The proof is a witness's reaction, the arc runs from shame to pride.
- Raise order value with the calendar.Black Friday, Holiday, Mother's Day, Prime Day and Labor Day each got closer ads and a pre-lander; the offer structure underneath stayed the same.
7 · Your KPIs and targets
This is the ladder from HQ/SCALING-CHECKLIST.md, with your 2026-09-27 rulings applied (Meta only up to $100k/day; no new channel or new country as a growth step; a new WHY only above $50k/day). The rung is set by one number: R7 = revenue per day over the last 7 days on orders containing the hero product. The right-hand column places Smooche's stages on the same ladder by ad volume only, since we do not know their revenue.
| Rung (R7) | The job | Targets to hit before moving up | Watch weekly | Smooche at this volume |
|---|---|---|---|---|
| $0 → $1k/day | Find one ad + page pair that pays | One ad at breakeven ROAS on 2+ separate days after 3× target CPA of spend · no funnel step below half its bar · CVR × AOV ≥ 200 on the lead page · 20+ new creatives a week, ported from proven donors | Meta ROAS vs breakeven, cost per add-to-cart, CVR, AOV | Summer 2025, ~15 ads/week, products being tested |
| $1k → $5k/day | Fill the lead pocket | MER ≥ breakeven over 7 days · budget steps within 72 h of 3 days at breakeven · 30+ new creatives a week · 2+ winning bodies · 20%+ video · 2+ advertiser pages · 3+ offer arms with orders · post-purchase upsell and core email/SMS flows live | MER, creatives/week, winning bodies, video share | Sep–Dec 2025, ~80 ads/week, engine on |
| $5k → $20k/day | Full depth and cheap width together | 100+ new creatives a week, 6+ opener families, closers 10–15% · video 50%+ across 2 video lanes · 3+ speakers and 3+ advertiser pages each with 10%+ of spend · 3+ page types with spend · 3+ doors tried on their own pages, 2+ alive past day 5 · 2+ person skins on every live door · AOV ≥ 1.5× single price | Tapped test weekly; doors alive; pages carrying spend | Dec 2025–Jun 2026, ~95–160 ads/week, foundation takes over |
| $20k → $50k/day | Many doors and person skins | 250+ new creatives a week · 10+ advertiser pages · 6+ doors tried, 3+ alive · 10+ live door × argument pairs and skins · a payer page (buyer is not the user) · second SKU PDP live · top 3 pages each have a twin | Pocket capacity, marginal ROAS per ad set | Jul–Aug 2026, ~260 ads/week, peptide serum added |
| $50k → $100k/day | New WHY, second SKU, languages | 400+ new creatives a week · a new population with its own WHY · 2+ SKUs each 15%+ of revenue · 60-day revenue per buyer ≥ 1.5× first order · R7 ≥ $100k held 14 days = done | Revenue share by SKU, cohort value | Peak weeks only (Aug 2026: 289–434 ads) |
Check 0 runs at every rung: R7 must print by command, a fresh daily ad ledger must exist, paid orders must carry their click, and breakeven ROAS and the cost-per-add-to-cart bar must be written in the offer brief. Until those hold, no budget step and no "tapped" call. Kills and new creative at held budget still go ahead.
The weekly dashboard, in one row per offer
| Number | Why it matters | Smooche reference |
|---|---|---|
| R7 and MER vs breakeven | Sets the rung; decides budget steps | unknown (no revenue data) |
| New creatives launched this week | The main lever; ad life is ~1 week | 15 → 80 → 160 → 260 → 430 |
| Share of those that reached breakeven on 2 days | Tells you whether new ads still win (the "tapped" test) | not observable from outside |
| Video share of new creatives | Video reaches placements statics do not | 42% → 74% |
| Live doors with their own page, and how many survived day 5 | Width on the same argument | 1 → 5 on the hero over 9 months |
| Advertiser pages carrying spend | Auction room and account safety | 3 → 10 in a week |
| Bridge pages with spend, by type | Yield per click | 109 built; 4–6 carry most traffic at any time |
| AOV and post-purchase take rate | How much each click is worth | $39 single, $119 pack, bumps |
Sources
teardowns/smooche/PLAYBOOK.md, PAGES.md, BRIDGES.md, SCRIPTS.md (teardown of 2026-09-07) ·
raw ads research/adlib/ads_smooche_domain_all.jsonl (7,388 ids, 7,162 linking to smooche.com, quiz., pre.) ·
weekly series built by teardowns/smooche/model/weekly.py → weekly.json ·
comparison brands teardowns/resilia/PLAYBOOK.md, offers/sofyre-rip/research/PLAYBOOK.md ·
ladder HQ/SCALING-CHECKLIST.md · the three playbooks HQ/PLAYBOOKS.md.