Mapped against the standards official statistics live by — point by point
SkyMind is built on internationally recognised statistical practice — and this page shows exactly where, framework by framework, row by row.
Every statement on this page links to evidence you can inspect yourself — without asking us.
This page maps SkyMind against:
- SDMX — ISO 17369 — the data model of official statistics, with a live SDMX-CSV export
- European Statistics Code of Practice — all 16 principles, each with a proof link
- FAIR principles — findable, accessible, interoperable, reusable
- Gulf national frameworks — Saudi ODL, Dubai Data Law, Qatar open data
Verification cannot be stopped here.
✓ met — click it to open the proof · — N/A by design · ↑ beyond the standard
What we claim — and what we don't. SkyMind's data model, quality rules and publication protocol follow the frameworks below, and every row in the tables links to a live page, an API response or a document you can check right now. What we don't claim is certification — no certifying body exists for any of these frameworks: SDMX is an exchange model, the Code of Practice is a peer-reviewed self-commitment among statistical institutes, FAIR is a self-assessment. Where a principle governs data collection from respondents, we mark it N/A and say why. If you find a row that doesn't hold, email info@sky-mind.com — corrections are published, not buried.
1 · SDMX — ISO 17369, the data model of official statistics
SDMX (Statistical Data and Metadata eXchange) is the only ISO standard specifically for statistical data. It is how Eurostat, the UN, the IMF, the World Bank and central banks structure and exchange statistics. Its core model: an observation is one value, identified by dimensions (who/what/when) and described by attributes (unit, status, source).
SkyMind's warehouse is built on the same model — one row, one observation, one full identity:
| SDMX concept | SkyMind implementation | |
|---|---|---|
| ✓ | Observation | One row in the observation store: (unit, metric, year, value) — never aggregated away, never rounded. ~2.10M observations. |
| ✓ | REF_AREA dimension | Official geographic codes where they exist — NUTS-3 (2024 edition) for the EU family — and stable SkyMind unit ids elsewhere (Israeli municipalities, Gulf communities/governorates, where no international sub-national code system exists). |
| ✓ | INDICATOR dimension | Stable metric codes (gdp_pc, psa_population, cen22_dependency_ratio…) — full catalog per country at /api/v1/catalog. |
| ✓ | TIME_PERIOD / FREQ | Calendar year, annual frequency; history back to 2000 for the EU family, 2001 for Israel. |
| ✓↑ | OBS_VALUE | Source value at native precision. Rounding is banned by policy — it manufactures fake ties (see methodology). |
| ✓↑ | Metadata attributes | Per metric: source organisation, source dataset, direct source URL, normalisation method, official/derived flag, date loaded — exposed at /provenance and in the API catalog. |
And it's not just a mapping on a web page. The export API speaks SDMX-CSV — the same CSV representation of the SDMX model that Eurostat's own dissemination API uses:
Available on Pro and Education API keys (the same paid gate as all bulk export). The precise claim: an SDMX-aligned data model with live SDMX-CSV output.
2 · European Statistics Code of Practice — all 16 principles
The Code of Practice is the trust framework of the European Statistical System — the 16 principles Eurostat and every EU national statistical institute commit to. SkyMind is a re-user: we unify what the institutes publish, and hold that layer to the same discipline. The principles that govern data collection from respondents are marked N/A below — which is itself a guarantee: the platform is built purely on published official statistics, with zero microdata by construction.
| Principle | How SkyMind meets it | |
|---|---|---|
| ✓ | 1 · Professional independence | Nobody pays us for conclusions. We publish documented observations, not narratives; the composite index is descriptive, within-country, with an open formula. No sponsor, government or client can order a number changed — every number is pinned to its official source. |
| ✓ | 1bis · Coordination & cooperation | We consume national statistical institutes' publications verbatim and cite the exact dataset per metric — /provenance names the institute, the dataset and the direct URL for every figure. |
| — | 2 · Mandate for data collection | N/A — we collect nothing. Only statistics already published by official bodies enter the platform. No surveys, no scraping of private data, no microdata. |
| ✓ | 3 · Adequacy of resources | The pipeline is automated end-to-end: scheduled refresh scans, a public freshness registry with estimated next-update dates per source (on the data page), and idempotent loaders that re-run safely. |
| ✓ | 4 · Commitment to quality | Every load passes a validation battery: impossible-value checks (shares >100%), carry-forward-year detection, coverage-gap checks, synthetic-pattern detection. In July 2026 that battery caught and removed an entire layer of unverifiable legacy Gulf metrics — publicly, with correction notes on affected publications. |
| ✓ | 5 · Confidentiality & data protection | Structurally guaranteed: the platform contains only aggregated official statistics. No personal data, no microdata, nothing to protect or leak. |
| ✓↑ | 6 · Impartiality & objectivity | The same data is free on the public map for everyone — customers pay for workflow and bulk access, not for different numbers. Publication protocol requires "what the data cannot tell us" alongside every finding (Evidence series). |
| ✓↑ | 7 · Sound methodology | Fully open: min-max normalisation within country, explicit axis weights, a full-basis gate that refuses to score years with thin data. Read and reproduce it. |
| ✓↑ | 8 · Appropriate statistical procedures | Documented, reproducible loaders per source; every dataset change logged in the public API changelog; deletions are backed up, never silent. |
| — | 9 · Non-excessive burden on respondents | N/A — no respondents. Zero collection burden by construction. |
| ✓ | 10 · Cost effectiveness | The platform is built entirely on open official data — the most cost-effective statistical raw material there is. What we add is unification, not duplication of collection. |
| ✓ | 11 · Relevance | Metric selection is "maximum relevant, not maximum count": income, employment, demography, housing, education — the questions people actually ask of regions (see Academy). Bulk irrelevant series are deliberately excluded. |
| ✓↑ | 12 · Accuracy & reliability | Values shown exactly as published by the source, at native precision. Known limitations are stated per metric (registration-place effects, small-cohort volatility). Cross-checks against independent totals accompany major loads. |
| ✓ | 13 · Timeliness & punctuality | Data currency is published, not implied: per-source current-through year, last refresh date and estimated next update are on the data page — including when the next Eurostat, CBS or GASTAT release is expected. |
| ✓↑ | 14 · Coherence & comparability | The core product: one schema across 47 countries, official geographic codes (NUTS-2024 applied within days of source datasets switching), boundary changes handled without fabricating continuity — when a region's borders change, its old series is retired, never silently re-labelled. |
| ✓↑ | 15 · Accessibility & clarity | Free interactive map, free API tier, JSON / CSV / SDMX-CSV formats, English labels everywhere, per-figure provenance drill-down. No account needed to inspect any number. |
3 · FAIR principles — Findable, Accessible, Interoperable, Reusable
FAIR (Wilkinson et al., 2016) is the de-facto standard for open data in research. The self-assessment:
| Principle | Evidence | |
|---|---|---|
| ✓ | F — Findable | Stable canonical URLs per country and district; official codes (NUTS, ISO 3166) as identifiers; schema.org Dataset markup; sitemaps + IndexNow; searchable metric catalog per country via /api/v1/catalog; a machine-readable DCAT-AP catalog feed at /dcat.json — the vocabulary data.europa.eu uses — one dcat:Dataset per country with coverage generated from the live database. |
| ✓ | A — Accessible | Everything over plain HTTPS. The map, the per-country metric catalog (/map/catalog), provenance and the changelog need no key at all; the versioned API uses a self-serve free key (POST /api/v1/key/free, 100 calls/month); bulk export gates are documented and priced openly on /pricing. Metadata stays free even where bulk data is paid. |
| ✓ | I — Interoperable | JSON, CSV and SDMX-CSV outputs; NUTS-2024 and ISO 3166 vocabularies; qualified references — every metric carries its source organisation, dataset id and direct URL. |
| ✓ | R — Reusable | Per-source licence audit across all 47 countries (July 2026); licence and attribution documents ship inside every data bundle (including a copy of the Saudi Open Data License in Saudi bundles, as its share-alike clause requires); per-figure provenance; prescribed attribution formats passed through to customers. |
4 · Gulf frameworks — the same discipline, applied where it's hardest
The Gulf has no equivalent of Eurostat, so conformance there means something more demanding: reading each national framework and complying with it specifically. What we operate under:
| Framework | What it requires — and what we do | |
|---|---|---|
| ✓ | 🇸🇦 Saudi Open Data License v1.02 | Attribution + share-alike for derived datasets published onward. We attribute GASTAT per metric and ship a copy of the ODL inside every Saudi data bundle (its §5.1 requirement). Sources: GASTAT Census 2022, HIES 2023, Labour Force Survey, Services Statistics — via the KAPSARC data portal. |
| ✓ | 🇦🇪 Dubai Data Law No. 26/2015 + DLD terms | Dubai's open-data classification governs Dubai Statistics Center data; Dubai Land Department terms prohibit reselling raw DLD data. We read that clause and restricted our own product accordingly: only normalised, derived metrics (Work Products) are distributed — never raw DLD passthrough. Geography follows the official DSC community grid. |
| ✓ | 🇶🇦 Qatar Open Data Policy (CC BY 4.0) | data.gov.qa publishes under CC BY 4.0 — attribution required. Every Qatari metric names its source (Planning and Statistics Authority, Ministry of Justice) with the dataset link in /provenance. Only the 8 official municipalities carry statistics; the 21 named zones are shown on the map but never assigned fabricated numbers. |
| ✓ | 🇮🇱 Israel (for completeness) | CBS publications under the Israeli government open-data terms; 837 metrics with the compendium edition named per load. |
The full licence matrix (17 source frameworks, all 47 countries, verified against licence texts) was audited in July 2026. Conclusion: everything we distribute is redistributable with attribution — and where a source demands more (Saudi share-alike, DLD work-products-only), we comply with the stricter rule.
5 · Where we go further than the standards ask
The codes above set the floor. These are the places SkyMind's own rules are stricter — each one checkable:
Standards require documenting sources at dataset level. SkyMind answers "where does this exact number come from?" for every figure: source organisation → dataset → original value → normalisation → direct URL. Try any district on /provenance.
Values are shown at the source's native precision. Rounding manufactures fake ties between regions (88.722 and 88.679 both become "88.7") — so it's banned by written policy, not by habit.
Charts carry a date and a document number (SKY-DE-20260707-0412) — deterministic, reproducible, citable. A screenshot of a SkyMind chart can be traced back to the exact data state that produced it.
When the July 2026 audit found unverifiable metrics in our early Gulf layer, they were deleted — and every affected publication received a visible correction note. Whole country indexes were left honestly empty rather than backfilled. The public API changelog records every data change since.
A country-year gets a composite score only when its full data basis exists (the "full-basis gate"). Thin years, boundary-change gaps and single-metric axes are skipped — a blank cell over a plausible-looking fake, every time. That's why some map areas are grey. It's deliberate.
When NUTS-2024 redrew regions (Rīga's merge, Lisbon's and Norway's splits), old series were retired and new ones loaded from the source's own geometry-consistent recalculation — never re-labelled. A new code means new boundaries; pretending otherwise would be fabrication.
Every Evidence publication ends with its limitations — GDP by place of production, commuter effects, small-cohort volatility — before anyone can misuse the finding. Standards recommend clarity; our publication protocol makes the caveat a required section.
Certificates can't be bought here. Verification can't be stopped here. That's the trade we chose.