Lender Complaint Index

Methodology & limitations

• ·COUNTS ARE ALLEGATIONS, NOT FINDINGS: each count is the number of consumer complaints CFPB received and forwarded to the company for a response. CFPB does not verify the underlying facts before publishing a complaint, and a complaint being counted here does not mean the company did anything wrong. Companies may dispute, resolve, or close complaints as 'no fault found.'
• ·NO DENOMINATOR / NOT SIZE-ADJUSTED: raw complaint counts are NOT normalized by number of customers, accounts, cardholders, loan originations, or assets under management. Comparing raw counts across issuers of very different scale (e.g. JPMorgan Chase vs. Pentagon Federal Credit Union) is misleading on its own -- larger institutions will mechanically show higher raw counts even if their complaint RATE (complaints per customer/account) is lower. This dataset does not include a customer-count or account-count denominator; treat raw counts as volume indicators only, not risk-adjusted comparisons.
• ·TAXONOMY CAVEAT (verified live against the API for this pull): CFPB's top-level 'Product' taxonomy changed in 2021 (e.g. 'Credit card or prepaid card' was split into 'Credit card' and 'Prepaid card') and has been revised further since. The product string requested for this dataset, 'Payday loan, title loan, or personal loan', is STALE and returns ZERO results as of this pull -- CFPB has renamed this category to 'Payday loan, title loan, personal loan, or advance loan' (the 'advance loan' addition captures earned-wage-access/cash-advance-style complaints, confirmed present as a sub-product bucket: 'Earned wage access'). This dataset uses the CURRENT, verified taxonomy string throughout, both as the query parameter and as the key in each company's byProduct object.
• ·hits.total IS NOT A FILTERED COUNT: on the geo/states sub-endpoint, the top-level 'hits.total.value' field always returns the size of the ENTIRE CFPB complaint database (17,142,190 records at the time of this pull) regardless of which product/company/date filters are applied. It must never be read as a filtered result. The correct filtered count for each query is 'aggregations.product.doc_count'.
• ·WINDOW BASIS: the date window (2025-07-01 to 2026-06-30) filters on date_received (when CFPB logged the complaint), not necessarily when the underlying incident occurred.
• ·COMPANY-NAME VERIFICATION: all 10 CFPB-registered company name strings were verified directly against the API (each confirmed via a nonzero, size-plausible 'Credit card' query before being trusted for the other 3 products). All 10 strings matched on the FIRST attempt with no fallback/variant substitution required, including the three the task flagged for extra verification (AMERICAN EXPRESS COMPANY, U.S. BANCORP, PENTAGON FEDERAL CREDIT UNION).
• ·TRUE ZEROS VS UNAVAILABLE: a 0 count is reported only where the company's name string was independently confirmed valid (via a nonzero Credit card result) and the specific product query itself returned a clean, fully-empty aggregation (not an error). Two cells could not be retrieved at all and are recorded as null (not 0): Capital One Financial Corporation and Bank of America, National Association, both for 'Payday loan, title loan, personal loan, or advance loan'. In both cases the exact CFPB-registered legal name combined with the current (longer) product taxonomy string produces a query URL that exceeds a hard, client-side ~252-character limit enforced by the fetch tool used in this environment (confirmed by systematic length bisection: 252 chars succeeds, 253 fails, independent of content). This is a tool-side constraint, not a CFPB API or data limitation -- the CFPB server itself was never confirmed to reject the equivalent longer URL. No shortened or alternate company-name variant (e.g. a narrower subsidiary entity) was substituted for these two cells, because doing so would silently change which legal entity is being measured and break consistency with the other 3 verified products reported for each of those companies. These two values should be treated as 'not retrieved', not as zero or low-complaint evidence.
• ·SUB-PRODUCT / ISSUE BREAKDOWNS ARE TOP-1 ONLY: 'topSubProduct' and 'topIssue' report only the single highest-volume category within that product x company cell for the window; they are not exhaustive breakdowns.
• ·SINGLE LEGAL ENTITY PER ISSUER: each company's counts reflect complaints logged against exactly one canonical CFPB-registered parent-entity name (listed in cfpbCompanyName). Large financial groups sometimes operate multiple distinct legal entities in CFPB's system (subsidiaries, predecessor names, servicing arms); complaints logged against a differently-named affiliate are not included here.
• ·Two products are structurally near-exclusive to certain issuers and a 0 there is expected, not an error: American Express does not originate vehicle loans/leases or (materially) mortgages, so low/zero counts in those categories are consistent with its business model rather than a broken query.
• ·All queryUrl values in this file are the exact, real URLs fetched to produce the paired count; each was independently verified to return a well-formed CFPB API JSON response before being recorded.

Author: Credit Defense Hub Editorial Team. Pending named subject-matter review — we do not display an expert-review badge until a real named reviewer exists. Corrections: via our contact page, recorded with dates rather than silently overwritten.