Skip to content
Banking RegistryEvidence-backed financial product data

How the registry works

Every published fact needs a paper trail.

Banking Registry separates what an institution officially discloses from what applicants and customers observe. Those are different kinds of evidence, and the site never presents one as the other.

Pipeline

From source to static page.

  1. 01

    Discover

    A regulatory, official, community, forum, or transcript source is identified and classified by what it can establish.

  2. 02

    Capture

    The raw artifact is archived immutably with its URL, retrieval time, content hash, media type, and source-use metadata.

  3. 03

    Structure

    Rates, fees, deadlines, issuer roles, eligibility, and unknown values become separate typed facts. “VERIFY” in a spreadsheet cell becomes a null plus an unverified status, never a number.

  4. 04

    Review

    New, conflicting, unverified, and high-impact claims are reviewed by a person. Automated extraction only proposes; it never publishes.

  5. 05

    Publish

    A claim publishes only when it is accepted, verified, backed by accepted evidence, and free of open conflicts. The static site is rebuilt from that projection.

  6. 06

    Monitor

    Each field class has a freshness expectation. Promotions recheck weekly; membership rules monthly; regulatory identity on regulator release.

Evidence states

What each label on this site means.

Confidence is claim-specific, not a permanent score attached to a website or creator. It considers source class, first-person versus second-hand evidence, recency, corroboration, and human verification.

  • Verified

    Accepted evidence and a current verification event exist.

  • Stale

    Verified once, but past its field-specific freshness window. Shown, flagged, and queued for recheck.

  • Conflict

    Two sources disagree, or an official rule contradicts observed outcomes. Both remain linked until resolved or contextualized.

  • Observed

    A dated community, forum, or transcript observation. Useful for bureau pulls and limits; never an issuer guarantee.

  • Not captured

    The registry has no accepted claim for this field. It is not a zero and not a “no.”

Evidence boundaries

Official terms are not community outcomes.

Sources are classified by what they can establish, not by a single universal reputation score.

Regulators and datasets
FDIC BankFind and NCUA data establish institution identity, charter, insurance, and financial records.
Official disclosures
Issuer pricing pages, agreements, membership pages, and benefits guides establish APRs, fees, eligibility, and application routes. CFPB terms surveys are a baseline, not the sole current-terms source.
Observed outcomes
Dated first-person reports, redacted evidence, and timestamped transcripts establish bureau pulls, starting limits, reconsideration, and reporting behavior. Published only as typed aggregates with sample size and window.
Research notes
Spreadsheet research is a lead, never a published fact. Every imported row keeps its workbook, sheet, row, and original text, and stays quarantined until accepted.

Privacy and honesty rules

If the source does not establish a term, the registry says so. It does not guess.

  • Numeric fields contain numbers, never prose. Units and date semantics are explicit.
  • Actual issuer and program family are resolved or visibly unknown. Personal membership never implies business availability.
  • Community records are anonymous by default; public pages use bands and aggregates, and small cohorts are suppressed.
  • The registry never collects Social Security numbers, account numbers, credentials, full birth dates, or unredacted credit reports.
  • No affiliate placement. Ordering is by the value you sort on, or by name.
  • Static output is rebuilt from the verified database projection; nothing is edited in a page template.