Living area is LIV_AREA in one system, SFLA in another, GrossLivArea in a third. Same house, three names — and before any model runs or any record moves, somebody rebuilds that crosswalk by hand. The profession has standards for data quality and standards for AVMs, but no standard for what the data is called, so every jurisdiction pays for the same mapping work over and over.
This session will do something about it in the room. Attendees receive a draft core schema of roughly forty fields — the attributes nearly every valuation model needs — and the panel’s job is to let the audience tear it apart, field by contested field. The draft is built on a plain naming convention the room can argue with immediately: fields carry a domain prefix that says what kind of thing they describe, so locational attributes read loc_street_type and loc_distance_to_water rather than arriving as forty unrelated column names. That is not a stylistic preference. It is the ISO/IEC 11179 pattern — object class first, representation term last — and it is the same move ISO 19152-4:2025 makes when it prefixes every valuation class VM_ to guarantee the name is globally unique.
Then the harder question: nobody knows which fields data science will demand in five years. The panel examines how healthcare (HL7 FHIR) and mortgage finance (MISMO) solved that with a core-plus-extensions architecture — a stable, versioned core, new fields entering as extensions until adoption earns them promotion.
And the question nobody in this profession has answered yet: what does the schema have to look like for something other than a human to read it? A REST API endpoint and a large language model tool call can perform the identical operation — fetch the comparable sales for this parcel — but they are described to the caller in different ways, and a field named LIV_AREA is a problem for both. The panel closes on what a schema owes an automated consumer, and why the answer to that question is the same answer that makes conversions cheaper today.
The panel spans public-sector modeling, client jurisdictions, research science, a national valuation regulator, and vendors, and the deliverable is deliberately permanent: a proposed IAAO working group, sign-ups taken on the spot, to carry the schema from straw man to standard. You leave with a marked-up schema worksheet you can hold against your own CAMA extract Monday morning.
Keywords: Data Standards & CAMA Integration, Advanced
