Published methodology
The K-Beauty Plug Demand Index — Methodology
Version 1.0 · Published · Next scheduled review: .
This is the public methodology for a metric we publish. It is versioned and dated so anyone can check what we claimed, when, and on what basis. If we change how the Index is calculated, the version number changes and the previous version stays online.
What the Demand Index is
The Demand Index measures what US K-beauty shoppers show interest in on kbeautyplug.com — which brands, categories, products, price points and retailers people look at, compare, and click through to buy.
It is built from first-party behavioural signals on a neutral price-comparison site. We sell no products, we take title to no inventory, and no brand or retailer can pay to change its position. There is no paid tier of the Index. See our honesty contract for how that neutrality is enforced.
What the Demand Index is not
We are stating these limits up front because a metric you cannot check is not worth citing.
- It is not a measure of US market share or US sales. It measures interest on one website. Our audience skews toward shoppers who are actively price-comparing, which is not the same as the general US buyer.
- It is not a sales figure. We observe intent — views, comparisons, click-throughs to retailers — not transactions. We do not receive purchase confirmation from retailers.
- It is not seasonally adjusted. We began collecting on 15 July 2026. Until we have observed a full annual cycle we cannot separate a genuine trend from a seasonal one, and we will not describe anything as “rising” or “falling” on that basis.
- It is not comparable year over year until July 2027 at the earliest.
How it is calculated
Every qualifying interaction is recorded as an event, tagged to the entities it relates to (brand, category, product, price band, retailer), and weighted by how much buying intent it demonstrates:
| Signal | Weight | What it means |
|---|---|---|
| Concierge route click | 8 | Acted on a specific recommendation |
| Retailer click-through | 5 | Left for a retailer to buy |
| Concierge query | 5 | Asked for something specific |
| Price alert set | 4 | Committed to wait for a price |
| Price comparison viewed | 3 | Actively evaluating where to buy |
| Wishlist add | 3 | Saved for later |
| Search | 2 | Directed intent |
| Product view | 1 | Passive interest |
For each entity we sum the weights over the period, then normalise to a 0–100 score where 100 is the highest-scoring entity in that group. Scores are relative within a period and within an entity type. A score of 60 does not mean “60% of anything.”
We publish the underlying event count (n) alongside every score. A score without an n is not a claim we have made.
Signals we deliberately keep separate
Three retailer measures exist and we never merge them, because they answer different questions:
- Retailer demand — a shopper chose that retailer and clicked through. This is preference.
- Best-price win rate — that retailer offered the lowest price in a comparison where at least two retailers were shown. This is price competitiveness, and it is our algorithm's judgment, not a shopper's.
- Coverage gap — comparisons where we could only show one retailer. This measures our catalog completeness, not any retailer's performance.
Combining these would let broad catalog coverage masquerade as shopper preference. We do not do it.
Suppression rules
We do not publish thin cells. As of v1.0:
| Entity | Status | Minimum to publish |
|---|---|---|
| Price band | Published | 20 events |
| Category | Published | 20 events |
| Brand | Top 10 only, n disclosed | 20 events |
| Product | Suppressed | 50 events |
| Retailer demand | Suppressed | 50 events |
| Best-price win rate | Published | 100 contested comparisons |
| Ingredient / concern / intent | Suppressed | 50 events |
Suppressed dimensions are collected but not reported. They will be published when they clear the floor, and we will say so when they do.
Exclusions and hygiene
- Internal, staff and known-automated sessions are excluded before any calculation.
- Percentage changes are reported in whole numbers or as rank movements. We do not publish changes to a decimal place, because our current sample does not support that precision.
- Where a single day or referral event accounts for an outsized share of a period, we say so in the report rather than smoothing it away.
- We describe findings as occurring “on The K-Beauty Plug,” never “in the US,” until panel size justifies the wider claim.
Publication schedule
| Date | What we publish |
|---|---|
| From November 2026 | Directional movers — category and price band only |
| From February 2027 | First quarterly Demand Index report, including top-10 brands |
| From August 2027 | First year-over-year and first seasonal commentary |
Until November 2026, the Index is used only in direct conversations with individual brands about their own products, where the alternative available to that brand is no data at all. It is not offered as an industry benchmark and should not be cited as one.
Corrections
If we publish a number we later find to be wrong, we correct it in place, date the correction, and leave a note saying what changed. We would rather be checkable than impressive.
Questions about the methodology: hello@kbeautyplug.com