Illustrative input columns
loan_number,address,city,state,zip
L-10025,11323 Astoria Dr,Charlotte,NC,28262
L-10026,5512 Pine Bend Ln,Tampa,FL,33617
L-10027,1987 Amber Way,Houston,TX,77084
Illustrative output fields
- Consensus value and range: one number plus low/high and confidence for fast screening.
- Geo overlay fields: flood zone, geo adjustment percentage, and key risk factors.
- Source-specific values: individual AVM readings where available for auditability.
- Stable identifiers: your original loan number or property ID carries through unchanged.
How low-coverage rows are handled
- If a property does not meet minimum source coverage, the row is flagged instead of silently dropped.
- Blank cells stay blank. The file still exports cleanly for downstream models.
- Where the product logic refunds tokens for low-quality outcomes, the refund is tied to the actual row-level outcome rather than forcing a failed batch rerun.
Why this matters
- Acquisition teams can sort and filter immediately instead of manually cleaning broken exports.
- Servicers can defend why a property was escalated or skipped.
- Engineering teams can ingest the CSV without having to special-case every missing field combination.