Original research
What changed in this personal-loan dataset
Read the latest changes first, then the earlier versions.
Each entry says what changed and why. A content update does not change the date of the source facts.
This 17-program sample is not the whole market and cannot predict your offer. CC BY 4.0 covers Rung-owned work, with third-party material excluded. This working report is public. Publisher review and third-party source clearance remain pending; no completed reviewer approval is claimed.
September 9: clearer findings and a separate APR appendix
Corrected the SoFi funded period on the advertised-versus-funded finding page to March 15, 2025–March 15, 2026. The previous page text incorrectly said February 16–March 15, 2026. The source record, 15.89% average, 11.90% later floor, and 3.99-point subtraction did not change.
Rebuilt the starting-APR guide around findings and borrower decisions. Added a separate loan-versus-card benchmark appendix with its own dates and limits; it does not expand the Collection 3 observations. Publisher review remains pending; no independent expert approval is claimed.
The dataset grew in 3 controlled steps
Scroll across to see all columns.
| Version | Date | Change | Reason |
|---|---|---|---|
| Collection 1 | 2026-09-05 | 5 structurally different programs | Test the public shelf-to-funded crosswalk |
| Collection 2 | 2026-09-05 | 6 more programs and borrower orientation | Expand fee, creditor, and funded-population evidence |
| Collection 3 | 2026-09-05 | 6 banks and credit unions plus state matrix | Compare traditional institutions and expose jurisdiction gaps |
The complete page and download package is built locally
The V4 candidate contains 41 routes, a 10-page report, data exports, a ZIP, page records, claim records, and repeatable builders.
A new snapshot must flow through every surface
- Recheck changing provider terms and source conflicts.
- Rebuild all scenarios from the stored inputs.
- Recreate the 867-row state file.
- Update HTML, PDF, CSV, JSON, schema, and citations together.
- Record material corrections without overwriting prior history.
More about this research
Cite the version you actually used.
A later provider update should not silently change what an older dataset said.