How the data is gathered, checked, and ranked
Every page on ChainCosts is built from a fact record, not from a writer's memory. Facts come from official provider pages, carry their own source link and verification date, and no page is published until it passes a fixed set of quality checks. This page describes the whole pipeline.
Last updated: 2026-07-11
Where facts come from
Fees, limits, supported regions, and product terms are extracted from providers' official published pages: fee schedules, pricing pages, help-center articles, and legal terms. Each extracted fact is stored with the exact source URL and the date we last verified it against that URL — and both appear on the published page, in the Sources section and next to the figures themselves. Secondary sources (press, blogs, forums) are not used as the basis for any figure.
Quality gates before publication
A page is published only if it passes every one of these deterministic checks. There is no editorial override — a failing page goes back to the pipeline, not onto the site.
- Length. The page must carry enough substance to answer its question; thin stubs are rejected.
- Near-duplicate detection. Pages are fingerprinted and compared; a page too similar to an existing one is rejected rather than published as filler.
- Banned phrases. Hype and guarantee language ("guaranteed profit", "risk-free", pressure phrasing) fails the page outright.
- Keyword bounds. Term frequency must stay within natural-language bounds — pages stuffed with a repeated phrase are rejected.
- Digit citation. Every number in the text must trace to a cited source fact or to a documented calculation over cited facts. A number with no source fails the page.
- Structured data validity. The machine-readable markup embedded in the page must parse and match the visible content.
- Geo-compliance. A provider is only shown in a market where its own published terms say the service is available; region-restricted services are excluded from editions where they cannot legally be offered.
How verdicts are computed
Comparison verdicts are the output of fixed rules applied to the fact records, computed the same way for every provider pair:
- The lower published base fee wins the cost row.
- The wider documented regional availability wins the coverage row.
- Rows are compared on published figures for the same scenario (same asset, amount tier, and payment method) — never a promotional rate against a standard one.
- When the published figures are equal, the row is shown as a tie, not broken by hand.
Commission relationships are not an input to any rule — see how we earn. If a verdict looks wrong, the fix is a data fix: correct the fact, and the verdict recomputes.
Update cadence
Source pages are re-checked on a rolling schedule, with frequently changing fee schedules checked more often than static terms. The per-fact verification date on each page tells you exactly how fresh each figure is. When a re-check finds a change, the affected facts are updated and every page built from them is regenerated and passed through the quality gates again before republication.
Report an error
If a figure on ChainCosts disagrees with the provider's official page, report it through the contact page with the provider name, the page URL, and the official source link. Confirmed corrections trigger an immediate re-verification and rebuild of the affected pages.