I built NationsHelm to make fragmented geopolitical data structured, comparable, and reconstructable.

Geopolitical data is scattered across dozens of institutions, coded in incompatible ways, and rarely dated where you can see it. Two numbers about the same country often can't be compared, and when one is wrong there is usually no one to tell and no record that it changed.

NationsHelm structures that data into one comparable frame, marks where every value came from, and keeps one name accountable for it. The point is not more numbers. It is numbers you can check.

Sources are selected on three criteria: they must be public, verifiable, and institutional — the World Bank, IMF, ILO, Transparency International, V-Dem, GDELT, UCDP, UN DESA, and their peers. A raw value is clamped to a documented domain and mapped to a score, or derived from other sourced inputs. When an input is missing it is rendered “—” and excluded from the blend — never filled with a synthetic placeholder or a fake zero. The full scoring model, source list, and per-metric formulas live on the methodology page.

Every value on the site carries one of four origin marks, so its provenance is always visible:

The line between a measured number and a written judgment is drawn in the open. Deterministic values are the ratings. AI writes analysis about them, always labeled, and it never crosses back into a score.

Every piece of analysis on the site is signed one of two ways, never ambiguously:

When a value is wrong, we correct it and log the correction. The log is public, entries are dated, and nothing is deleted from it. You can flag any figure two ways: email corrections@nationshelm.com, or file a per-value report from the corrections page.

Predictions resolve the same way. Every published prediction carries a resolution date and resolves in public — on the card that made it. A miss stays on the record, stamped MISSED, not quietly removed.

The scoring model is versioned with semver. A formula, coverage floor, composition, or clamp-domain change bumps the version and adds a dated line to the changelog — so every score can name the exact version that produced it, and any change to how a number is made is public and dated.

Methodology v2.5.0Owner: Caro B.

Self-funded. No ads. No data sales. NationsHelm takes no advertising and does not sell user data. Revenue, when it comes, is from subscriptions and licensing — stated on the pricing page.

Conflicts of interest: none to disclose. There are no affiliations, holdings, or arrangements that could bias a score toward any country, leader, or party.

Model vendors:the one AI feature in production — the leaders' Analyst Outlook — is generated with Anthropic's Claude Opus 4.8. It is labeled on every card and never blended into a deterministic score.

For editorial questions, press, and anything requiring the editor directly: caro@nationshelm.com. It is a real, monitored inbox. To flag a specific figure, use corrections@nationshelm.com or the corrections page.

Citing NationsHelm in research or journalism? The citation page has ready-to-use APA, MLA, Chicago, and BibTeX formats, and the rule that matters most: cite the underlying primary source for the raw number, and NationsHelm as the platform that compiled, scored, and dated it.