I counted the same shelf seven times.
The wall never moved K-beauty · Olive Young ranking shelves, 479 brands over 7 weeks · adjacency · public data only

01Seven weekly snapshots of a shelf

The data first. Each weekly snapshot of the ranking pages catches 1,655 to 1,686 brands and about 9,500 products. The seven-week union holds 2,044 brands — 68% appear all seven weeks, 9% appear once and vanish.

The floor of the shelf turns. Of the products ranked in week one, 25.9% are gone from the rankings six weeks later. Brands are different — 86.7% of week-one brands still rank somewhere six weeks on. Products churn; brands stay. It's the same picture as the food registry in the previous piece, except here the rotation is visible week by week.

This piece measures not the rotation but the walls. The analysis covers the five categories where crossing is most visible — skincare, masks, cleansing, sun care, and makeup — which held 479 brands in the final week.

02Open doors and a closed wall

Every beauty company has sat through the meeting where someone says, "shouldn't sunscreen be next?" Once skincare finds its footing, somebody always brings it up. So let's count how that meeting actually ends on the shelf. These are the brands appearing in two categories' rankings in the same week (week of 2026-08-10):

Adjacency is the ruler this piece builds: the number of brands actually present on both shelves ÷ the number you'd expect if presence were pure chance. At 1.0 the two shelves have nothing to do with each other. Above 1.0, there's a road brands actually travel. Below 1.0, brands cross less than chance itself — that's not a road, it's a wall.

The table splits cleanly in two. Among skincare, masks, cleansing, and sun care, all six pairs run at 1.17–1.44 — the doors are open. Every pair involving makeup runs at 0.34–0.50 — below half of chance. Of the 138 brands on the makeup shelf, 39 also appear anywhere in the skincare cluster: 28.3%.

That is the hole this piece found. Expansion plans tend to price the two identically — same budget logic, same success metric, same meeting. The shelf has been treating them as different events all along.

03Not one week's accident

One cross-section could be sampling luck. So the same calculation, all seven weeks:

I'll be honest — when I first built this table I expected the numbers to wobble week to week. They didn't. Across seven measurements, the six open pairs and the four makeup pairs never once crossed into each other's range. The open side never dips below 1.1; the closed side never climbs past 0.5. The products on these shelves churn every week, and this structure doesn't budge — which means it's a property of the market, not of the sample.

I counted the movement, too. A brand appearing for the first time in a new category's ranking — a shelf-level "expansion attempt" — happened 51 times across the five judgeable weeks. Ten-odd events a week across four-hundred-some brands: expansion is a rare event, not routine. And of those 51 attempts, 29 were still on the new shelf the following week — 56.9%. Nearly half don't survive their first week.

04What to actually do

If you've been reading this with your own next category in mind, put it against this ruler. The arithmetic is the one above — brands present in both your category and the target, divided by what chance alone would produce. The threshold is 1.0. "Chance" isn't anyone's opinion; the normalization builds that line in.

Two choices in this calculation, stated with their counterfactuals. The denominator's N is all brands in the five categories' rankings — widen it to all of Olive Young and home-goods brands inflate every adjacency; narrow it to one category and there's nothing to normalize against. And co-presence counts within a single week only — count it across the seven-week union and one-week cameo entries read as standing crossovers, inflating the rate (the union-based crossover does come out a few points higher). This piece takes the lower, conservative reading.

05Whose hands hold the data behind the shelf

This is as far as counting from outside can go — who sits on which shelf. The next questions — who bought, and did the person who bought the skincare also buy the same brand's sunscreen — aren't in any ranking. That data accumulates on the side that owns the shelf. And Korea's competition authority has already written down, twice, what that arrangement looks like.

Reading the KFTC's 2023 sanction briefing, I stopped at one passage. From 2017 through 2022, Olive Young provided suppliers with sales information on their own products through its internal system — and charged for it, at roughly 1–3% of net purchases, collecting ₩170 billion over six years as "information processing fees." The KFTC sanctioned this in 2023 under the Large-Scale Retail Business Act. Read it plainly: brands paid, for six years and ₩170 billion, to see the records of their own products being sold.

The Coupang decision (2024) contains the same asymmetry running the other way. In developing its private-label products, per the decision, Coupang "extracted products using the product information suppliers enter when registering to supply, and the internal information accumulating in its own online mall as products sell." That sentence is a statement of fact, not the violation — what was ruled illegal was pushing those private-label products up with manipulated search rankings and 72,614 employee-written reviews (a ₩162.8 billion fine). Put together: the sales data accumulates on the shelf's side, and the shelf used it to pick which products to compete against its own suppliers with.

The law does not correct this asymmetry. I checked the statute: the Large-Scale Retail Business Act contains no clause obliging a retailer to share sales data with suppliers. What it contains is the opposite direction — Article 14 prohibits the retailer from demanding suppliers' business information. The EU went the other way: Digital Markets Act Article 6(10) obliges gatekeeper platforms to give business users free, effective access to the data generated by their activity, and Article 6(2) bans the platform from competing against them using their non-public data.

Which is why this series keeps returning to the same spot. The final ingredient that separates an expansion from a new entry — do our customers buy both categories — exists only for brands with their own channel and their own data plumbing. Everyone else can see the market average (this piece's tables) and must decide about themselves with their eyes closed.

06What this shelf cannot show

Rankings are not sales. Olive Young's "best" ordering has no published formula. Every sentence here stands on "exposed on the shelf," never "sold the most."

The 51 expansion attempts are shelf events, not launches. An old product newly entering the ranking counts too, so the attempt count and survival rate carry that noise — the direction holds, the magnitudes deserve slack.

Nothing outside the five categories was counted. Expansions into body care, hair, or fragrance don't register here. How open those doors are can be measured the same way.

Customer-level truth is not computable from outside. Cross-purchase, cannibalization, genuinely new customers — the numbers §04's re-verification row demands all require an order ledger and customer IDs. As §05 laid out, where that data is piling up right now is exactly the problem.

07What this piece is really saying

Getting on a shelf and widening your shelf are different events. Seven weeks of counting showed the skincare cluster trading brands at 1.2–1.4× chance — open roads — while every path into makeup sat below half of chance the entire time: a wall. Yet the single word "expansion" covers both, and most plans price them identically. Before you pick your next category, measure its adjacency — above 1.0 it's an expansion, below 1.0 it's a new entry no matter what the deck calls it. Then check whose servers are collecting the cross-purchase data you'll need to finish that judgment.