1. Keep the source record attached to the parcel
Avara’s data path records ingestion sources, runs, snapshots, and manifests before the information becomes a canonical parcel record. In Vantage Parcel Audit, reviewers can see the source, source record, source hash, valuation effective date, model-generated date, and the time the underlying record was fetched.
That evidence distinguishes the record actually used by the workflow from a source that may have changed later. It also gives a reviewer a direct starting point when a parcel characteristic, assessment history, sale, or permit signal needs verification.
2. Tie every value to its model run and policy
A parcel value is not fully identified by a model name. Vantage retains the model version, run ID, production run identity, and parameter hash needed to connect the parcel to its sales window, feature set, calibration, clamps, and other run controls. Parcel Audit exposes that provenance in its source-details appendix.
At the run level, Model Audit organizes evidence into run identity, methodology, training evidence, scoring evidence, soundness checks, and feature attribution. This lets reviewers move from an individual value to the policy and run that produced it without reconstructing configuration from an analyst’s memory.
3. Explain the final value in assessor language
Parcel Audit follows the value from the source assessment through model factor, calibration, equity controls, rounding, and final equalized value. Its valuation explanation shows the source fields behind major drivers, their observed values, peer context, direction of influence, and dollarized effect on the final value.
The page also connects comparable and study-group context to the same value. Reviewers can see which evidence raised or reduced the result and whether the displayed effects reconcile to the stored model delta. That is more useful than a generic feature-importance chart because it explains the parcel under review.
- Source assessment and source hash
- Model factor, calibration, and final equalized value
- Observed source fields and peer statistics
- Model name, version, run ID, and parameter hash
4. Audit the run, not just the parcel
Avara Model Audit presents the stored methodology and the evidence that the run was ready to use. Reviewers can inspect baseline strategy, qualification rules, time-index treatment, calibration and clamp settings, training-row evidence, scoring coverage, data gaps, clamp hits, and soundness checks such as whether the current scoring run completed and feature contributions are available.
The same surface can export the complete run evidence and a feature summary. That gives model-risk, leadership, and audit teams a portable record of the run behind the roll instead of a collection of screenshots from separate tools.
5. Carry evidence into review and appeal preparation
Vantage Outliers assembles parcel evidence around the reason a record was flagged. The inspector can show required evidence fields, source signals, peer statistics, an evidence timeline, suggested checks, permit status, and the policy version used for reproducibility. The language explicitly treats the parcel as a review candidate that needs source verification—not as a conclusion about the property or owner.
Reviewers can export parcel evidence while Parcel Audit retains the value explanation, source details, comparable context, and model provenance. Together, those surfaces create the supporting context for an appeal-ready record and make routine quality review use the same evidence that would be needed under scrutiny.