Put every loan in one pool. Kadince is designed to hold your full portfolio (mortgage, small business, small farm, commercial, consumer) regardless of pre-identified eligibility. The same records are then viewed through CRA, HMDA, fair lending, and mapping lenses via filters — no separate databases, duplicate queries, or duplicate records.

Three ways to get loan data in, and the TXT shortcut is the easiest. A HMDA LAR pipe-delimited TXT file auto-maps on import with zero mapping work. CSV exports from an LOS/core require a one-time mapping template (saved per system; column order doesn't matter as long as row 1 is the header). API integration is available but your IT team must build the query. Kadince checks for and rejects duplicates when pulling from multiple sources.

Edit checks run live against the CFPB. Records are validated in real time against the Filing Instruction Guide — change a field and the check re-runs immediately. Validity errors must be fixed; quality checks can be marked resolved with a rationale comment that persists for auditors and examiners. The "fix it" link jumps straight to the offending field, and widespread issues (e.g., a missing credit score column) can be cleared in bulk via report mass-update or a data-tool re-import. Partial exemption can be set by year, which auto-flags exempt fields and pulls them out of edit checks while still letting you track them internally.

AI workflow checks records against their source documents. Before edit checks even come into play, a custom AI workflow can read attached loan docs and flag mismatches against field values (e.g., income entered as $75K but $57K in the document). You define the instructions and knowledge sources, and it works on any loan type, not just HMDA.

Geocoding confidence is customizable and fixable. Any loan with an address is geocoded to FFIEC tract and demographic data, with a confidence score and thresholds you set. A low-confidence dashboard surfaces problem records; you can retry a different address or drop a pin to restore 100% confidence and lock in the correct tract, MSA, county, and demographics. Note that dropping a pin will overwrite the address, so it's pin or address, not both.

Submissions go straight through to the CFPB. Quarterly, cumulative (Q2 automatically includes Q1), or annual filing — no downloading files or emailing them elsewhere. Kadince runs one final check for syntactical/validity and unresolved quality errors before sending, and generates a CSV working file listing the record IDs and why each was flagged.