Which public system a figure came from — OpenDataPhilly payroll, OPA, L&I — not an unlabeled number.
Update cadence varies by source; that variance is stated, not hidden behind a single "as of" date.
TaxAssessmentIQ names the specific comparable sales behind a recommendation — not just a percentage.
A neighborhood score or a flagged pattern is only useful if the underlying records behind it are visible too.
What area or date range a figure covers, so it isn't mistaken for a citywide or all-time claim.
If an assessment looks fair or a portfolio looks clean, ALKARTIS says so — not every check should end in a recommendation to act.
Explainability isn't a feature added on top of the product — it's a constraint on what gets built in the first place. If a conclusion can't be traced back to a specific, checkable record, it doesn't ship as a conclusion.