Narrative insights

CFPB tags every complaint with an issue/sub_issue from a fixed taxonomy. This clusters recent complaint narratives with KMeans (unsupervised, no taxonomy involved) to see whether anything CFPB doesn't already have a name for shows up — a new scam pattern, for instance. Site-wide across all companies, not per-company (see why in How it works).

Clustering run: 2026-09-08 · narrative window 2026-06-17 to 2026-06-30 · 2000 narratives.

Clusters marked unreliable below overlap heavily in the underlying t-SNE projection, consistent with CFPB's own redaction placeholders (xx, 00) polluting the term vectors — a known, not-yet-fixed issue (see the project's CLAUDE.md). Treat their labels as unverified until that's fixed.

Unauthorized account opening △ unreliable

334 narratives · xx, 00, debt, credit, collection, xxxxxxxx, payment

Debt validation requests under FDCPA

129 narratives · debt, collection, validation, alleged, original, reporting, creditor, fdcpa

Unauthorized transactions and misrepresentation △ unreliable

863 narratives · card, 00, money, credit, xx, funds, payment

Unauthorized transactions and errors △ unreliable

411 narratives · xx, year, 00, payment, card, credit, merchant

Account closures and fraud restrictions

55 narratives · app, cash, xx, money, transactions, 00, did, support

Inaccurate Credit Reporting Claims

208 narratives · credit, reporting, information, report, accounts, inaccurate, 15, reported