Skip to content
⌘KStart free
AI InfrastructureDefenseQuantumAll studies →

Methodology — Revenue-Geography Overlap Chips (and the Ambiguity Policy)

Transparency is a feature. This page is the published policy behind the revenue-geography chips in the pair drill-downs — what a chip means, what it can never mean, and exactly how ambiguous data is handled. The policy ships with the feature: if the chips are visible, this page governs them.

Two of your names can share an economic driver the price history alone won't name — both earn a slice of their revenue in the same part of the world. Where both companies explicitly tagged revenue for the same curated region in their latest 10-K, the pair drill-down shows a chip naming that region, with each company's tagged share of its own revenue.

A chip is a fact about disclosures: "both filers tagged revenue in this region." It is not a risk score, not a correlation estimate, and never a return forecast.

What the data is — and is not

  • Source: each company's latest 10-K XBRL segment disclosures (the same data behind the Segments panel). Labels are shown as tagged in the filing's XBRL — the member names, de-camel-cased — never re-worded by us.
  • Coverage is ~22.5% of the universe (the live denominator is shown beside the chips). Companies that report as a single segment disclose no geographic split and are absent by design. That absence is missing data, not evidence of diversification — a name with no chip may share every driver with your book. Foreign filers (no 10-K) have structurally zero coverage here.
  • Customer-concentration flags (the SFAS 131 "customer ≥ 10% of revenue" disclosure) are not built: customer identities are absent from our ingest and XBRL anonymizes them. We do not substitute a guess.

The curated buckets

Every stored label maps to at most one bucket, or to an explicit no-data class. The buckets, in display order:

United States · Canada · Latin America & Caribbean · North America (as-tagged aggregate) · Americas (as-tagged aggregate) · Europe · EMEA (as-tagged aggregate) · Middle East & Africa · Greater China · Japan · Asia-Pacific — other · Asia-Pacific (as-tagged aggregate)

Curation calls we made deliberately (not oversights):

  • Greater China means explicit China / Hong Kong / Macau tags only.
  • Taiwan is bucketed under "Asia-Pacific — other", never under Greater China — a data-bucketing choice, not a political statement.
  • Turkey is grouped under Middle East & Africa (business-convention EMEA grouping). Russia and Ukraine are grouped under Europe (geographic).
  • US territories (Puerto Rico, USVI, Guam) count as United States.
  • Named US regions of a US filer ("Northeast", "Midwest", …) count as United States.

The ambiguity policy

  1. "Asia Pacific ≠ China, ever." An as-tagged aggregate ("Asia Pacific", "EMEA", "Americas", "International") never matches a finer bucket, and a specific tag never matches an aggregate bucket — no imputation in either direction. A company tagging "Asia Pacific" and one tagging "China" do not chip together: we cannot know how much of the first company's Asia-Pacific revenue is Chinese, so we refuse to guess. The cost is missed matches; the alternative is fabricated ones.
  2. Complement tags carry no geography. "Non-US", "International", "Rest of World", "Other Countries" and kin go to an explicit non-specific class — disclosed per name, matched with nothing.
  3. Ambiguous two-letter codes are quarantined. Historic records dropped the XBRL namespace, making codes like CA (Canada? California?), DE (Germany? Delaware?) and IN (India? Indiana?) genuinely undecidable — both readings occur in real filings. These are shown in an explicit ambiguous class and never bucketed. Records rebuilt by the namespace-aware ingest resolve this at the source (state members are suffixed "(state/province)"), after which bare codes are countries.
  4. Anything not in the curated map stays unmapped — disclosed as such. The map is versioned with the code; it never guesses.

Why nothing here is ever summed

Filers tag overlapping regions: "US" inside "Americas", "China And Hong Kong" inside "Asia Pacific" — sometimes in the same filing. Summing overlapping tags would double-count revenue, and summing risk bars over overlapping driver tags would manufacture exposure. So:

  • bucket figures are never added across buckets;
  • when several tags fall into one bucket, we use the largest single tag — a conservative floor under any overlap structure, never an inflated sum;
  • shares are each company's tagged revenue over its own reported fiscal-year total. If the fiscal years can't be matched, or a tagged segment exceeds the reported total (inconsistent denominators), the share abstains with the reason — the chip still names the shared region.

Reading the chips honestly

  • A chip means both companies disclosed revenue there — it does not size a common risk, and two chips do not mean "twice the overlap."
  • No chip can mean: no shared geography, or no data (the coverage line and the per-name no-data reasons distinguish these — read them together).
  • Shares are of revenue, not of profit, and reflect the latest annual filing only.