ZKH US Equity☆

Today

— Today's value on the same definition as the history: the live screener figure where the definition matches, otherwise the newest stored snapshot (the cell says which month).

Own median

— The median of this name's own monthly point-in-time values for the selected window — the robust middle of its own history, not a peer median.

Own range

— Position inside this name's OWN range for the window: the rule runs min→max, the shaded block is p10–p90 (p25–p75 when the window is too short for tails), the grey tick is the own median and the dot is today. An arrow at an edge means today sits outside its own historical range.

%ile

— Percentile of today's value within this name's own monthly distribution for the window — mid-rank for ties, the same convention the peer panel below uses. Coloured cheap/strong versus rich/weak by the metric's own orientation.

z

— Population z-score of today's value inside that same own-history distribution: (value − mean) ÷ σ. Blank when the distribution has no dispersion to speak of.

Sector med.

— The same-month median across this name's sector cohort in the same archive, at the newest month available — the cell's tooltip gives that month and how many names stood behind it.

mo · distinct

— Monthly points · DISTINCT values. The ratio metrics are annual figures re-read every month, so the distinct count is the effective sample size — it is what a band would actually rest on. A series needs 12 months; a band needs 60 months AND enough distinct values to cover the years it spans; p10/p90 need 24 months in the window.

P/E

— as-of close ÷ latest knowable FY diluted EPS (split-rebased); the live screener P/E uses the identical definition on today's close

ROE

— net income ÷ average total equity, latest knowable FY

ROIC

— NOPAT ÷ average invested capital, latest knowable FY

Net Margin

— net income ÷ revenue, latest knowable FY

12M Momentum

— trailing 12-month adjusted-price total return at the snapshot date

EPS Growth 3Y

— 3-year CAGR of diluted EPS over knowable FYs (split-rebased)

Sales Growth 3Y

— 3-year CAGR of revenue over knowable FYs

Market Cap

— shares outstanding × close, USD-normalized

Every snapshot stamped reconstructed was rebuilt by re-computing this name's metrics from only the facts an observer could have read on that date: a fact enters a month only once the annual filing that covers it (10-K/20-F/40-F) had a real SEC filed_date on or before that month-end. Per-share figures are split-rebased before the ratio is taken. Snapshots with no basis stamp are forward-captured ("live") and are labelled as such above, never folded in silently.

Restatement vintage: the availability GATE is honest (a fact enters a snapshot only once its covering annual filing's real filed_date precedes that month-end) but the VALUE stored is the latest restated one — an observer in 2017 may have seen a since-revised figure.

Survivorship: only names present in today's ingest exist in the archive, so a cohort median excludes companies that were delisted or acquired along the way.

Order-of-magnitude artifacts were dropped at write time by per-metric plausibility bands (P/E kept within 0.5–3,000×, ROE/ROIC ±50, net margin ±100, momentum −1 to +500) — real extreme tails are kept, so a band can legitimately be wide. A TRAILING P/E in particular explodes whenever earnings trough (CSCO's FY2018 tax-charge EPS of $0.02 gave a real 2,340× reading in 2019), so read the median as the robust statistic and the p90 as 'this is what a trough month looked like', not as a rich-valuation month.

P/E and EPS growth are reconstructed for US-domestic 10-K filers only — a 20-F/40-F foreign private issuer's ADR trades in USD while its EPS is reported in the native currency, so no multiple is reconstructed rather than a wrong one.

Sector medians use the cohort's CURRENT classification, not a point-in-time one (the same caveat the exposure and manager-sector surfaces carry) — a name that changed sector is counted in today's sector for every month it appears.

Gates: a series needs 12 monthly observations, a percentile band 60, p10/p90 inside a window 24, and a sector month 5 contributing names. Each row above states which gate it missed.

Effective sample size, not month count: the ratio metrics are annual figures re-read monthly, so the same value repeats between filings. Nothing is ranked inside fewer than 4 DISTINCT values, and no band is quoted unless the series carries at least 6 distinct values per decade of span — a "139-month range" resting on four fiscal years would misstate its own evidence. Both counts are in the last column.

P/E is the only series here with a price leg, so it is the only one that can be wrong from a per-share basis mismatch: when the live pipeline has flagged this name's share basis as changed, or the reconstructed level falls below what any listed equity trades at, the whole P/E row is withdrawn in favour of its reason rather than published. Every other series is a total-dollar ratio — currency- and split-neutral. 140 monthly point-in-time snapshots, 2014-12-31 → 2026-07-14, reconstructed by ingest.pit_reconstruct. 1 of 140 snapshots carry no pit_meta basis row and are reported as forward-captured ('live'), not reconstructed. No series for: P/E, EPS Growth 3Y, Sales Growth 3Y, Market Cap — each row states why.