# Methodology and checks

Snapshot: 20261015T151411149784Z, retrieved October 15, 2026. The source CSV used for analysis has SHA-256 f7a5026104a897ebe327c80e48b9186aa168ce4c7d306d0f6829b89d36b8edf9. The narrower downloadable candidate has its own hash in manifest.json.

## Acquisition

Use the City issued-building-permit catalog and public GIS layer linked in Sources.md. Request the layer metadata, record count and complete object-ID list. Fetch attributes in batches of at most 500 IDs, reject truncated responses, reconcile returned IDs to the complete list, and check that source editing metadata remained unchanged during capture. Preserve dated responses before transforming them. A single API response may not contain the whole dataset.

## Permit-record transformations

All 15,912 source rows are retained, including CountyFire and repeated Application strings. OBJECTID identifies rows within this capture; Application is a source string, not a verified project key. Type and Sub_Type are administrative labels. Status_Date is parsed as month/day/year and grouped by calendar month. No dates are imputed, rows deduplicated or work labels independently reclassified.

The public candidate retains eight source fields plus snapshot_id, source_url and review_flags. It excludes addresses, contractors, descriptions and parcel values. Missing-field counts measure exact empty strings, not semantic completeness. Placeholders and incorrect nonempty values require separate review.

Monthly/type/subtype tables count source rows and reconcile to 15,912. The July comparison uses separate denominators of 1,785 Residential and 269 Commercial rows; another 38 July rows are outside those two types. Chart clicks filter the record table; they do not change the fixed July chart.

## Housing-unit series

This is a separate 2020-2025 series, not a join to the 2026 permit browser. City totals sum USER_PER_UNIT by YEAR. HCD APR Table A2 selects Sacramento, reporting year and BP_ISSUE_DT1 within that calendar year, then sums NO_BUILDING_PERMITS, defined by HCD as units. The six annual sums match. Reporting year alone produces different totals. Download housing/annual_units.csv for both selected and unfiltered comparison fields.

## Checks and limits

Capture, schema, hashes, counts, IDs, dates and aggregate reconciliation were checked. Browser filters, CSV export, chart selection and desktop/mobile layouts were tested locally. These checks establish reproducibility and interface behavior, not source correctness, geographic completeness or completed construction. No month has verified complete coverage; CountyFire scope remains unverified.

Keep corrections tied to snapshot and source field; preserve original captures. This is a local review candidate, with no agency endorsement or claim of live jobsite reporting.