Methodology

The atomic figure is a layer score — a 0–100 reading for one layer of a supply chain (for example, “indium phosphide wafer” or “optical transceiver module”). We call the current formula cpx-2.0. It is a weighted combination of three components measured over the tracked players in that layer:

The layer score is the weighted mean of the available components, scaled to 0–100. When a component cannot be measured — most often customer-concentration fragility, which depends on voluntary disclosure — its weight is dropped and the remaining weights are renormalized, so the score is always a proper weighted average of what we actually observed.

Layer scores roll up into the higher-level figures the same way each time:

• A chain score is the mean of the top-3 scored layers in that chain.
• A sector index is the mean of that sector’s scored chains.
• The site-wide Chokepoint Index is the mean of the top-5 scored layers across the whole map.

Rollups only ever average scored layers or chains (see Section 04). A layer we decline to score contributes nothing to the figures above — it is not treated as a zero.

The inputs are all public filings retrieved from the U.S. Securities and Exchange Commission. Four sources feed the map:

SEC XBRL company-facts supply each player’s latest annual revenue. We read the structured company-facts feed and take revenue from a fixed candidate set of four us-gaap tags — RevenueFromContractWithCustomerExcludingAssessedTax, Revenues, RevenueFromContractWithCustomerIncludingAssessedTax, and SalesRevenueNet — restricted to annual periods (10-K / 20-F filings, roughly 330–390 days). Companies migrate between these tags over time, so selection is recency-primary: the entry with the most recent fiscal-year end wins, across all four tags. When two tags report the same fiscal year, the tag order above breaks the tie; the latest-filed entry breaks any remaining tie. We take one figure from one tag — never a blend — and record the tag, fiscal-year end, filing date, and accession number as the citation.

SEC Financial Statement and Notes Data Sets feed the disclosure graph we use to discover supply-chain membership — which companies name each other as customers or suppliers in their filings.

EDGAR full-text search surfaces additional candidate companies by keyword across the filing corpus.

The customer-concentration screen reads disclosed major-customer revenue shares from 10-K and 10-Q filings; it supplies the fragility component in Section 01.

Refresh cadence. Scores are recomputed quarterly, after filing season, so newly filed annual and quarterly reports are reflected. Each score carries the “as of” date of the run that produced it.

These numbers describe disclosure structure. They are not the things people sometimes assume:

• Revenue concentration is not market share. The HHI is computed over the players we track in a layer, not over an entire market. A high reading means revenue is concentrated among the tracked names — not that those names control the market.
• Disclosed customer shares are dated filing facts, not current figures. A top-customer share is whatever a company reported in a specific filing period. It is not a live number and may be stale.
• None of this is investment advice or a return prediction. The Chokepoint Map makes no alpha claims — no figure here forecasts a stock’s return or is offered as a reason to buy or sell. This is a standing editorial policy, stated plainly.

We would rather show nothing than show a figure built on too little disclosure. A layer is scored only when both of these hold:

• at least 3 players have a usable revenue figure, and
• those scored players cover at least 50% of the layer’s tracked players.

When either condition fails, the layer is marked “insufficient disclosure” and shows no score. It is left out of every rollup rather than counted as a low or zero reading.

Foreign filers without U.S. XBRL revenue and private companies still count as players — they are part of the map — but they are excluded from the computation because we have no comparable revenue figure for them. The coverage ratio makes that exclusion visible: a layer with many untracked or foreign players will show lower coverage, and may fall below the 50% bar.

Companies reach a layer through one of three routes:

• Seed curation — an analyst places an anchor company in a layer to start the chain.
• Disclosed-relationship discovery — the filing disclosure graph surfaces companies that name each other as customers or suppliers.
• Keyword discovery — full-text search over EDGAR surfaces additional candidates by subject.

Discovery is a shortlist, not an admission. Every entry is reviewed and approved by a person against filing evidence before it appears on the map, and the route it came in through, the evidence link, and the approval date are recorded as provenance.

• cpx-2.0 (July 20, 2026) — the initial published formula described on this page.

The version string is stored on every computed figure. Any change to the component weights, or to how a component is measured, bumps the version so a score can always be traced to the formula that produced it.