Most composite indices fill a gap and move on. This one withholds the composite and publishes the reason, and it distinguishes between two reasons that are usually collapsed into one. That distinction is the methodological contribution of this work and it is the page worth reading.
Recording all three as geopolitical data-accessibility exclusions would be accurate for Taiwan and wrong for Hong Kong and Singapore, whose absence is a catchment resolution limit that no access would resolve. One label over three different causes is the kind of claim a reviewer checks once and then distrusts throughout.
SOURCE_RESOLUTION
The source measures at a resolution that does not resolve this territory. The value is absent for a measurement reason, and no amount of access would produce it.
SOURCE_POLITICAL_COVERAGE
The source's territorial schema does not carry this entity as a separate entry. The value is absent for a coverage reason that is political in origin, not measurement error and not missing data.
Every known limitation of this vintage, derived from the index at build time rather than written afterwards.
U-01 · HIGH · coverage
3 of 35 states carry no composite: HKG, SGP, TWN. Water stress is absent and coverage falls to 0.70 against the declared 0.75 threshold. The absence has two distinct causes, published per state as withholding_cause.
Effect. Those states cannot be ranked or compared on composite.
Remedy. Scoped as permanently out of coverage against Aqueduct 4.0 rather than pending. HKG and SGP carry SOURCE_RESOLUTION: no WSE value exists at source for a catchment of that extent, so no amount of access produces one. TWN carries SOURCE_POLITICAL_COVERAGE: the source's territorial schema does not publish a separate entry. Reopening this requires a different water-stress source with city-territory resolution and its own territorial schema, which would be a new axis input and not a gap-fill.
U-02 · HIGH · configurational
The published fsQCA is not a calibration of the published axes. Five independent violations are recorded as D-103: identical raw values with divergent fuzzy membership, an inverted order between GBR and DEU on WSE, a near-constant condition present in every sufficient path, solution statistics that are internally inconsistent, and an outcome variable that contradicts the index it is meant to explain.
Effect. Pathway memberships, solution statistics and necessity results in vintages through v8.4 are not derived from the axes published beside them. The reported consistency of 1.000 on all three paths is an artefact of the near-constant condition, not a result.
Remedy. Re-run complete, and the necessity claim is no longer withheld pending a computation. The computation was run and the claim did not survive it. REE returns necessity consistency 0.7033 with Relevance of Necessity 0.5766 and coverage of necessity 0.5763, holds above 0.5 in 29 of 32 cases, and scores 0.7268 against the negated outcome, which is higher than it scores against the outcome itself. The gate returns TRIVIAL. Publish the four measures rather than a bare consistency figure, and withdraw the necessity claim rather than restating it at a lower number.
U-03 · MEDIUM · fidelity
Fidelity is computed for 3 of 35 states (DEU, ESP, GBR).
Effect. Structural Lack is unavailable for the rest of the sample, so no fidelity ranking exists.
Remedy. Extend the Bhattacharyya computation to the full sample or state the three-state scope wherever fidelity is cited.
U-04 · MEDIUM · measurement
Availability-zone counts are missing for 27 of 35 states, so the ODI Herfindahl is weighted one unit per active cloud region rather than by capacity.
Effect. Operator concentration is measured by region count, which treats a one-zone and a six-zone region alike.
Remedy. Complete intra_region_fault_domains in cloud_regions_master.csv; the capacity-weighted value is already published where the source records it.
U-05 · MEDIUM · scope
FPE, ESE are published per state and enter no composite (D-044).
Effect. The headline number omits the only fully component-audited axis (ESE) and the platform-exposure axis (FPE).
Remedy. Derive a weight from something, or keep stating the exclusion wherever the composite is presented. Do not assert a weight.
U-06 · MEDIUM · scope
No semiconductor axis exists. The February 2026 roadmap named semiconductor capacity as the Infrastructure Pillar. Scoped as a declared roadmap item, target vintage v9.0, 2026-12-31, contingent on identifying a 35-state fabrication-capacity source. Until that date the roadmap claim is stated as pending, not delivered.
Effect. Compute-fabrication dependency is unmeasured. The roadmap claim is scoped rather than unsupported: it carries a target vintage, a date and the condition that would let it ship.
Remedy. Identify a 35-state fabrication-capacity source before 2026-12-31 and derive a weight, or record the item as retired on that date. An undated roadmap item is indistinguishable from an abandoned one.
U-07 · LOW · lineage
The v8.0 ordinal ODI carried four manual overrides (NLD, HKG, KOR, SGP), each one step below its own stated rule, recorded in no export.
Effect. Any analysis reproducing the v8.0 ODI from its documented rule will disagree on those four states.
Remedy. Superseded by D-045. Retained here so the v8.0 discrepancy is explainable rather than mysterious.
U-08 · LOW · measurement
CSE composes EMODnet and SubmarineMap at 0.6/0.4 for EU states and uses SubmarineMap alone elsewhere.
Effect. The axis is not constructed identically inside and outside the EU, so cross-region CSE comparisons carry a method difference.
Remedy. State the split wherever CSE is compared across regions.
This index is an MCP server. Any AI client that supports MCP can read it directly and get an answer bound to the hash below, rather than to a screenshot of this page.
{"mcpServers": {"qesis": {"type": "http", "url": "https://qesis.eu/mcp"}}}
# then ask: qesis_get_integrity