SciRisk 100

Methodology

How every number in this index is produced, and what it is not.

One engine

Every ESCRC figure in the SciRisk 100 is produced by the same production ESCRC engine that powers the public ESCRC Calculator — SciRisk ESCRC Model v5.4.2. There is one implementation of the model mathematics in the codebase and both products call it. Given identical inputs, the calculator and this index return identical results; an automated test asserts that bit-for-bit on every build.

What ESCRC is

Value-at-Risk: the portfolio loss not exceeded with the chosen confidence level. Read as the ESCRC risk capital — the loss the firm should be prepared to absorb in a bad year. Model-implied and intended for screening and ranking, not a calibrated capital figure.

Mechanically: for each supplier the model draws minor (operational) and major (natural-hazard or geopolitical) disruption events over a 365-day horizon, multiplies the resulting disruption days by the supplier's unrecoverable share of daily revenue exposure, sums across the network, and repeats 100,000 times. ESCRC is the 95% quantile of that loss distribution. Major events are correlated within a geography group through a one-factor Gaussian copula (rho = 0.35), so a regional shock hits every supplier in that region together.

Deeper supplier tiers are attenuated by echelon buffer factors and capped by flow conservation: a tier's propagated exposure can never exceed the tier below it. Cumulative disruption days are capped at the horizon.

Risk Capital Intensity

Risk Capital Intensity is ESCRC divided by the company's reported revenue. It is not a separate model — it is the engine's output over a declared company input. Absolute ESCRC scales with company size; intensity does not, which is what makes it comparable across a $20bn company and a $600bn one.

SciRisk Supply Chain Risk Grade

Companies are graded on Risk Capital Intensity across five bands. Preliminary SciRisk 100 classification — methodology subject to validation. The thresholds below are round-number cuts placed against the observed distribution of this dataset so that each grade carries a meaningful share of the universe. They are not calibrated to external loss experience, to any capital standard, or to any rating scale, and they are not part of the ESCRC methodology.

Company inputs

The model needs, per company: revenue, cost of goods sold and gross margin; and per supplier: tier, geography, product, focal-firm-attributable purchase volume, minor- and major-event rates, the corresponding recovery times, and the unrecoverable fraction of daily revenue.

Financials are curated from company reporting (Company annual reports and Form 10-K / 20-F filings), on the basis stated in the dataset: Most recently reported full financial year as at the data date (FY2024 or FY2024/25 depending on the issuer's calendar). Revenue, cost of goods sold and gross margin are taken as reported; figures are rounded to the nearest USD 0.1bn and converted at the issuer's own reporting rate. Supplier networks are curated from Supplier relationships and tier assignments curated from company supplier-responsibility disclosures, published supplier lists, procurement filings and reported sourcing relationships. Purchase volumes are declared as a share of the focal firm's cost of goods sold.

Per-supplier model parameters are not researched company by company in this release. They are filled from production artefacts under one uniform rule for all 100 companies: minor- and major-event rates come from the ESCRC model's own calibrated industry and geography prior buckets (priors 2026.09-5, whose derivations are documented in the ESCRC Methodology Specification), and recovery times and revenue-at-risk fractions come from the median of the shipped calculator preset networks within the same industry bucket.

What that means for the reader. Two companies in this index differ because of their sector mix, supplier geography, tier depth, supplier concentration, the share of cost they place with external suppliers, and their revenue-to-cost structure — not because one has been researched more closely than the other. For the six companies where the calculator ships a researched network, running it on the same engine gives between 3.63x and 35.84x the index figure. That spread is not a measure of research alone: every company in this index is priced pro-rata, by each supplier's share of cost, while those six calculator presets declare a per-supplier criticality and are priced on it, which prices a concentrated bottleneck far above its share of cost. The parameter layer and the severity formula both differ, and the formula is generally the larger term. Each of those six pages shows its own comparison; none of them is used to adjust anything.

Aggregation across companies

The index headline, and every sector total, is a plain sum of independently-calculated company-level ESCRC figures: the engine has no portfolio- or system-level model, and produces no such figure itself. Aggregate ESCRC is the sum of independently-calculated company-level ESCRC estimates. It should not be interpreted as unique system-wide (or sector-wide) supply-chain risk capital: company supply-chain exposures overlap — many suppliers in this dataset appear in more than one company's network — and the model does not net out or de-duplicate that overlap when figures are summed.

Reproducibility

The dataset is deterministic. One pinned seed (20260821), the engine's own seeded generator, and every version that can move a figure recorded on the output: engine 5.4.2, priors 2026.09-5, dataset 2026.09-23, calculated 2026-09-21. Re-running the build on an unchanged repository reproduces the same figures exactly. Each company also carries a fingerprint of the inputs that reached the engine, so a moved number can be attributed to inputs or to the model.

What this is not

SciRisk ESCRC is an analytical estimate generated using the SciRisk Economic Supply Chain Risk Capital methodology. It is not a credit rating, investment recommendation or statement of financial condition. Figures are model-implied and intended for screening and comparison, not as calibrated capital requirements.

Beta / preliminary dataset. Company inputs are analyst-curated from public annual reports and public supply-chain disclosure. ESCRC itself is computed by the production ESCRC engine — only the INPUTS on this page are preliminary.