Ask any caseworker why a household's rent changed and they will open Elite, OnSpring, or SACS. The screen displays an income figure, an effective date, and a calculated rent. Case closed. Then somebody, perhaps a hearing officer, legal aid, or a HUD reviewer, says "show me," and the agency learns that the screen is a witness, not the record.
PHA software is good at formal coherence. It catches broken date sequences, blank required fields, and transactions that do not fit the workflow the vendor imagined. It is much worse at detecting a thin file, stale verifications, an unsigned 9886, or income that was countable on paper but not in fact. A 50058 that passes HUD's data edits proves only that the record was internally consistent. It does not prove the documents in the cabinet supported the entry. "Our system indicates" is a weak sentence because it summarizes nothing more than what the agency wrote down for its own convenience.
The hierarchy that staff and software both forget is that the tenant file is the proof layer and the software is the production layer. The signed 9886, pay stubs from the correct period, SSA letters, EIV reports, and the rest of the documentary record are what support the judgment. The software stores the abstraction that follows the judgment. When that order reverses, and the screen is consulted as if it were evidence, neat entries start receiving the respect that belongs to sound casework. Scanning weak files into a document management system does not fix this. It produces searchable weak files.
The remedy is dull and durable. Define what each file must prove. Build SOPs that say what evidence must be present before a transaction moves. Run QC on a sample before HUD or a tenant's attorney does. Train staff to reach the judgment before they touch the keyboard. None of this is exciting, which is partly why vendors prefer to sell storage. Our operational assessments, SOP development and file digitization work all begin in the same place: when somebody says, "show me," what can the file actually prove?