Banking fee data methodology

The Business Banking Cost Calculator and the per-provider fee pages are built on data extracted from the fee schedules each provider publishes on its own website. This page describes exactly how that data is collected, what has to happen before a number is published, and how corrections work.

Collection

A registry maps each provider to the canonical URL of its published fee schedule. A weekly pipeline fetches every source (HTML pages directly, PDF schedules via text extraction, bot-protected sites through a headless browser). The extracted text is hashed, and if a source is unchanged since the last run the stored figures are simply re-stamped as freshly verified. Changed content is re-extracted with strict structured output and validated against numeric sanity rules: maintenance fees must fall between zero and 100 euro per month, percentages within plausible ranges, no required field missing.

The two-fetch rule

A provider's figures are never published from a single read. The first valid extraction is held as a candidate. It goes live only when a second fetch on a different day either re-extracts to exactly matching figures or returns byte-identical content, which proves the source is stable. If the two extractions disagree on any figure, the provider goes to a manual review queue and nothing is published.

Cadence and freshness

Sources are refreshed weekly. If a provider's data has not been successfully re-verified for 60 days it is flagged "verification pending", shown without ranking figures, and excluded from the headline benchmark until the pipeline confirms it again.

Validation and review

Anything that fails validation, extracts with low confidence, cannot be fetched, or fails the two-fetch comparison lands in a review queue for a human decision. Nothing moves from the review queue to the live site without a person approving the change in a pull request. A number without a source URL does not exist on this site.

Corrections

Every publication and every subsequent change is recorded in a changelog with the date, the provider, the fields that changed, and the source document. Corrections are made by publishing a new entry, never by silently editing a figure.

Scope and limitations

The data covers standard published fees for entry-level business current accounts: maintenance, electronic transaction charges, cash lodgement, FX margin, and card fees where published. Negotiated pricing, waivers, introductory offers, overdrafts, lending, and merchant acquiring are out of scope. This is an indicative reference, not financial advice.

Last verified