Methodology · Binna

Seven commitments

Facts only. Every score rests on a fact with a public link: a retailer's listing page, an announcement, or a government disclosure.
A published, versioned formula. Any change to the formula gets a new version number and is announced one edition before it takes effect.
Open data. Every edition comes with a CSV download, licensed CC BY 4.0, so anyone can recalculate it.
Rank is not for sale. Sponsorship and payment never affect whether a brand is included or where it ranks. Sponsored content is labelled separately. Brands the operator has a commercial relationship with are excluded from every ranking.
Frozen editions. One edition a month. Once published, an edition is frozen and archived at a permanent link, and it is never quietly changed.
Public corrections. Brands and readers can submit evidence and ask for a correction. Corrections go into a corrections log with the date and the reason.
Stated coverage. Every ranking says which retailers were checked, when they were checked, and which ones cannot be checked automatically.

Independence and conflicts of interest

Rank is not for sale. This section states who is kept out of the rankings and how.

Who is excluded. Any brand the operator of Binna has a commercial relationship with, including brands it is in talks with, is excluded from every ranking. The rule applies whatever the brand's retail coverage.
Where the rule is applied. The exclusion runs in code when the public data is built, before any ranking is calculated. Excluded brands do not appear anywhere on the site, so they cannot be ranked. Nobody decides it brand by brand when an edition is made.
Sponsorship is a separate question. Sponsorship and payment never affect whether a brand is included or where it ranks, and sponsored content is labelled. See the sponsorship rules.
What is disclosed. This rule and the fact that the operator has such relationships, also stated on the About page. This page does not list the excluded brands by name.

Formula v1.0

North America Shelf Index and Canada Shelf Index

Score = the number of distinct mainstream retail banners that carry the brand.

Mainstream retail means beauty specialty retailers, mass retailers, pharmacies and department stores.
Banners are counted separately for each country. The Canada Shelf Index counts Canadian banners only.
Only brands that are on shelf count. Announced launches and online-only listings do not.

Ranking

Standard competition ranking (1, 2, 2, 4). Brands with the same score share a rank and are shown in alphabetical order.

New on Shelf

Brands that reached a mainstream North American retailer in the last six months, ordered by the month of first listing, newest first. Only listings dated to the month are included.

The weekly chart

The K-Beauty North America Chart on the homepage is the North America Shelf Index, published once a week. It uses formula v1.0 unchanged: brands are ranked by the number of distinct mainstream US and Canadian retail banners that carry them in store, with standard competition ranking.

When it is taken. Each week's chart is the shelf index as calculated on the first daily data run of that week (weeks run Monday to Sunday). Once taken, that week's chart is frozen and not changed. Listings checked later in the week count from the next chart.
Last week. The brand's rank on the previous weekly chart (normally last week's; a week with no chart is not filled in), or a dash if it was not on it.
Move. ▲ and a number: up that many places. ▼ and a number: down that many places. A dash: same rank. NEW: on the weekly chart for the first time, which can mean a newly added brand in our data as well as a new launch. RE: back on the chart after missing at least one week. On the first weekly chart there is nothing to compare with, so every brand shows a dash.
Peak. The best rank the brand has held on any weekly chart, this week included.
Weeks. The number of weekly charts the brand has appeared on, this week included.
Who is on it. Every brand with at least one mainstream in-store banner on record. The homepage shows ranks 1 to 10 with all tied brands, unless a tie at the cut-off is too large to list there; the full chart lists every brand.
Brands that leave the public data during the week. If a brand is removed from the public data after the chart is taken (for example because a commercial relationship begins, or a correction removes its listing), it is taken off the current chart straight away. The other brands keep their ranks until the next chart.
Weekly chart and monthly editions. The weekly chart does not replace the monthly editions. Each month's edition is still frozen on the 1st with a CSV download and remains the record to cite.

How listings are verified

For each brand and each mainstream retailer, we look for the retailer's own brand or product page, or a public announcement of the launch.
Automated checks run only where a retailer's robots.txt allows them, and never more than one request to the same site every five seconds.
Where automated checks are not allowed, listings are checked by hand each month, or brands send us the evidence and we verify it against the retailer's page before including it.
Every listing shows its source link and either the date we last recorded a check of it or the date it was reported in the news. The source link may be the retailer's page, a brand announcement or a news report. Reported listings count toward rankings unless the report says the brand is sold online only.
Each month we draw a random sample of 30 published listings, check each one against its source one by one, and publish the result, including the listings we got wrong, on the audit page. The audit does not change any score or rank.
Our track record counts every launch announcement we published, from the day it appeared on our site to the first evidence that the brand went on sale, under rules registered in advance, including announcements that did not come true. It does not change any score or rank.

Evidence levels and listing dates

Each brand and retailer pair carries the fields below. They are derived from the same public records as the rankings and are published in the CSV files.

Verified: We recorded a check date for this listing. The source link may be the retailer's page, a brand announcement or a news report. The date shown is the most recent check date we recorded.
Reported: The listing comes from a news report and we have not recorded a check of it yet. The date shown is the date of the report.
First seen: The earliest date in our records for this listing: the month it went on sale, the date of the first report of the listing or its announcement, or the check date we have recorded, whichever is earliest. For an announced launch, the planned month is not used. For a listing with no sale month and no report, first seen is the most recent check date we have recorded, which can be later than the day the product actually went on sale. Whenever first seen comes from a check date, it can move later after we re-check the listing.
Last seen: The most recent evidence that the brand was on sale there: the latest check date we recorded or the latest report of the listing, whichever is later. Empty for a launch that has only been announced.
Delisted: Set only from a public source that the brand has left the retailer, after we have checked that source. The delisting date is the date of that source. A delisted retailer does not count toward any shelf index. If newer evidence shows the brand on sale again, the listing returns to listed.

Dates keep the precision of their source: a year (2017), a month (2026-08) or a day (2026-10-11). When dates of different precision are compared, a year or month counts from its first day.

Editions frozen after these fields were introduced include two CSV files: the rankings, one row per ranked brand, with the number of verified and reported retailers behind its score; and the shelf events table, one row per brand and retailer, including announced, online-only and delisted listings. Both are frozen with the edition.

Shelf time on record

Each brand page shows how long the brand has been on shelf in our records: the number of days from the earliest first seen date to the latest last seen date across its listed retailers. Announced launches, delisted retailers and distributor announcements are not counted; online-only listings are. The count is conservative: a first seen date known only to the month or year counts from the last day of that month or year, and a last seen date known only to the month or year counts from its first day (this differs on purpose from the comparison rule above, so that shelf time is never overstated). It is the span of our evidence, not the brand's actual time on sale, and it can grow as new evidence arrives. Shelf time is shown for reference only and does not enter any ranking.

Hero product and key ingredients

Where we have checked it, a brand page names one hero product and the key ingredients listed for that product. The hero product is the first single product in the bestseller section of the brand's official site; sets, bundles, devices and items outside skincare, haircare and makeup are passed over. If the brand has no English-language site, the site has no such section, or we could not read it or that product's page, it is the top single product on the brand's page at Ulta or Sephora in that retailer's default order. Key ingredients are the ones named as key ingredients on that product's page (on the brand's site or the retailer's page we used) or in the product's name: the first three in the order listed. Each entry links to the page we read (the product page, or the bestseller list when that is where the ingredients are named) and shows the date we checked it. Formulas change, so the entry is only as current as its check date.

These are labels of what the brand says is in the product. We do not rate ingredients, efficacy or safety, and we make no clinical claims. Hero products and ingredients do not enter any ranking. Brands we have not checked yet show "Not recorded yet".

What we do not rank

Prices. Tracking them at scale would mean large-scale scraping of retailer sites.
Ratings or reviews. Nobody could recalculate them.
Any subjective "best of" selection.

Changes to this method

The current formula is version 1.0. Any change will be published here with a new version number one edition before it applies. Found something wrong? Submit a correction.