# Branded Civic Review Packet: analytical source receipts

Source: the uploaded 33-page **Praeva Analytics ALPR Audit Records Civic Review Packet**, review date September 2026. Page numbers below are the printed page numbers.

This is a project-provided analytical report. It is not the original VCSO response letter, policy, contract, Flock/Axon audit export, or Camarillo presentation. Reported agency positions remain agency representations. No public release of the complete uploaded PDF is assumed.

## Page 4 — request, agency response, policy and capability
The packet summarizes the September 14, 2026 agency response: Organization and Network Audit logs were produced; the response described these as the only available responsive information for most audit items. Purpose/reason was reported as omitted under an investigatory-records exemption. Moorpark user mapping was described as not tracked, with no responsive record.

The packet reports shared oversight roles, formal ALPR training records not maintained, and some configuration/retention records unavailable or not tracked. These are reported agency positions.

The contract summary lists broader network access, time-and-location searches, plate lookup, vehicle-attribute search and audit reporting.

## Pages 5–6 — audit rows and screening
An audit row is not one person, vehicle or investigation. Delivered and exact-distinct rows must be distinguished. Missing/unavailable fields are not numeric zero. March 2026 is partial. Candidate geography and organization classifications require reconciliation.

The non-California label profile screens frequency, persistence and requested scope. Labels may represent agencies, centers, task forces or export naming conventions.

## Pages 10–11; methods page 31 — peer comparisons
Page 10 explains why several comparison frames are retained: California sheriffs, a volume-matched activity band, a population band and nationwide sheriffs. The report describes high frequency and long-window measures in several frames and more mixed typical reach.

The four sample sizes printed in the supplied original graphic are 40, 4, 4 and 848 respectively. They are transcribed, not recalculated. The activity band is volume-matched, not function-equivalent. Membership/classification remain subject to reconciliation.

Percentiles are tie-aware and descriptive. Small cohorts are particularly sensitive to membership choices. The report explanation is now supplied; underlying peer-membership and metric tables are not.

## Pages 24–25 — NCRIC
Page 24 reports 83,555 delivered and 82,595 exact-distinct NCRIC-attributed rows across 27 active months. It describes median monthly rows falling from 7,421 in January–July 2024 to 1,181 in August 2024–February 2026, while median reported network reach rose from 214 to 755.

The report lists query construction, operations, export coverage and reporting structure as possible explanations. The final NCRIC activity band contains 17 labels and is volume-matched, not function-equivalent.

## Pages 28–29 — action and questions
The packet separates observed patterns, review questions and alternatives.

It calls for field-preserving audit records, criteria/reviewers/results, authorization/training checks, sharing/configuration/lifecycle documentation, and investigation of observed temporal changes. Its listed alternatives must be tested rather than assumed.

## Still missing
The PDF itself does not identify the requested California/local misuse cases or supply the exact official transparency/trust quotations. A subsequent source-recovery note identifies the California State Auditor precedent and attributes trust wording to the Camarillo presentation. See recovered-source-receipts.md for the verified official audit, attribution boundaries and remaining primary-receipt requirements. The per-state time series and raw peer tables are still not supplied by the PDF.