Half the revenue happens on someone else's broadcast.
Whose ledger holds the customer? Eight Korean supplement brands · home shopping, own storefront, marketplaces · public-data analysis

01What the companies disclosed

Start with the table. All three figures are the companies' own FY2025 filings.

Company A books nearly half of product revenue on television. Company F splits three ways, and the 6.2% line labeled TM is orders taken over the phone. Company D is overwhelmingly online.

None of that is a verdict. It's a business mix. Home shopping has always been large in this category, and given who buys these products, a live phone channel surviving into 2025 makes sense.

02Two different businesses share one cell

The problem is what each company means by "online." Their own words:

The domain in the second row is redacted because it would identify the company. Everything else is the filing's own wording.

The company's own storefront and third-party marketplaces sit in the same cell. Nobody is hiding anything — the disclosure format simply doesn't ask for that split. So in the income statement and in the annual report alike, the two live on one line.

A sale through your own store leaves you the buyer ID, the reorder interval, the abandoned cart, the banner they arrived from — all of it on your own servers. A sale through a marketplace leaves you a settlement statement. Same "online," entirely different residue.

Home shopping is starker. The name and number of the person who watched the segment and picked up the phone sit with the broadcaster. Telemarketing leaves its own trace somewhere else again. The revenue lands on your income statement, but the person who produced it lives in another company's database.

03This is not a mistake anyone made

Worth saying plainly: a high home-shopping share is not a failure. These companies are in that channel because it was the most efficient way to move product in this category. The same goes for marketplaces.

What it does is make return on spend structurally hard to count. Suppose an ad brings someone in. If they buy through home shopping, nothing in your own data tells you whether they bought again three months later. And if you can't see that, you can't establish whether the ad made money or lost it.

In this category that matters more than most, because repeat purchase is the business. Collagen, probiotics — nobody buys these once. Being unable to count reorders is more expensive here than almost anywhere else.

04The spread within one industry

Widen the lens and the channel structures diverge sharply. Here is FY2025 for eight brand companies.

Heavier advertising does trend with thinner margins — across the eight brands the correlation is −0.72. But it is far from clean, and the places where it breaks are the interesting ones. Company B spends 25.14% on advertising and turns a profit; Company F spends 11.72% and loses money. The heavy spender earned, the light spender lost.

If "spend less on ads" were the answer, that line would be much straighter. It isn't — which means how much you spend is not the deciding variable.

05The gap isn't missing tooling

One thing I wanted to rule out: maybe the companies with broken margins simply lack measurement tools. So I pulled the page source of each company's own store and counted the advertising and analytics scripts.

Company A's store carries seven — GA4, Google Tag Manager, the Meta pixel, Kakao, Naver conversion tracking, Criteo, and Google Ads. The most in the cohort.

A single company demonstrates that the count of tools and the presence of measurement are different things. The first piece in this series found the same shape in K-beauty — every pixel installed, and nothing above them doing the deciding. What's different here is that it shows up in the income statement.

And there's a simpler point underneath. At a company booking half its revenue on television, seven pixels on the web store can only see the smaller half.

06The auditor looked at the same seam

Reading Company F's FY2025 consolidated audit report, I stopped at the key audit matters. The auditor had designated one: the occurrence of revenue through the company's own online mall.

In the auditor's words: as the company diversified its sales channels to grow revenue, new transaction types including online sales were added, and among these, revenue recognition through the company's own online mall carries high inherent risk — so the risk of misstatement in the occurrence of that revenue was identified as a significant risk.

The audit opinion was unqualified. This does not mean anything was found wrong. It means the area carried elevated inherent risk, so the auditor sampled transactions against source documentation, and then issued a clean opinion. That is the only way this paragraph should be read.

But the direction is the same one. The audit firm noted the diversification of sales channels first, and then singled out revenue through the company's own channel as the part requiring separate verification. The seam we identified from the outside — two businesses in one "online" cell — is the same seam someone looking from the inside decided to examine on its own.

07What public data can and cannot show

Visible — channel revenue and mix, for the companies that disclose it; advertising and selling-commission expense in absolute terms and as a share of revenue; cases where the channel is written into the chart of accounts itself (Company D books a line item literally named broadcast sales commission); the type and count of tracking scripts on each store; whether membership tiers and loyalty points exist.

Not visible — repeat-purchase rate, contribution margin by channel, the incremental effect of advertising. Those require the ad accounts, the order ledger, and channel settlement data. And revenue from the own storefront alone — except that one turned out not to end at "not visible."

So everything to this point is a hypothesis about where to look, not a measurement. But even what can't be seen has an outline. The next section measures it.

08When a point won't come out, measure the interval

No company discloses its own-storefront revenue as an exact figure. But the ceiling and the floor that the filings permit can be computed. The quantity being measured is the owned share — revenue where a customer ID lands in your own ledger ÷ total revenue — and applying each company's written definition of "online" literally gives this.

The first thing that catches the eye: one of the three can be read. Company F defines "online" as its own mall alone, so its owned share is effectively public. Same disclosure format — one company legible, two not. It isn't the format that sets the resolution; it's what each company chose to bundle into one cell.

And the column that matters more than the values is the width. 6.2pt against 78.8pt — same format, a twelvefold difference in resolution. A narrow interval means even an outsider can move straight to the next question (repurchase). A wide one means attribution has to be settled internally before any number downstream means anything.

09So how should repurchase be reported?

The most common mistake in this category is reporting repurchase as a single number. At a company booking half its revenue on someone else's channel, the sentence "our repurchase rate is 32%" is almost always a statement about the storefront.

The owned share measured in the previous section belongs next to the repurchase rate every time. Without it, a reader cannot tell whether the 32% describes the whole business or a slice of it.

This is not about modesty; money turns on it. Repurchase feeds lifetime value, and lifetime value sets the ceiling on acquisition spend. Report the storefront number as the company number and LTV inflates — and you overspend by exactly that much.

10Three choices that decide this calculation

• Start home shopping and marketplace revenue at zero in any customer count. Do not fill the gap by inferring from settlement data. A settlement line is a transaction, not a person, so counting them as people invents customers who don't exist. Zero is not a wrong value; it is the honest notation for "unknown."
• Never use the storefront's repurchase rate as the company's. Someone who navigates all the way to your own store already has an attachment to the brand. Leave that selection bias uncorrected and repurchase is structurally overstated.
• When correction isn't possible, drop cross-channel comparison and read the time series within each channel. You cannot say whether storefront repurchase beats home shopping. You can say whether storefront repurchase improved on last quarter — the bias sits on both ends and cancels. For most decisions that is enough.

And there is one item worth negotiating out of a broadcast partner. It is aggregate, not personal data, which is what makes it winnable: the ratio of first-time to repeat orders per broadcast. No names, no numbers. That single ratio moves you from seeing none of the larger half to seeing which way it is moving.

11What this piece is really saying

One "online" line holds both the own storefront and the marketplaces, and that boundary decides whose hands the customer data ends up in. When a point estimate won't come out, measure an interval — the upper and lower bounds of the own-channel share, and the width between them. Narrow width: go straight to repurchase measurement. Over 40 points: attribution comes before any advertising conversation. And a repurchase rate quoted without its attribution share is a number nobody can place — the whole business, or a sliver of it.