First-time residential inspection pass rates, measured

For each permit, we take every inspection event of a given type whose result maps to a pass or a fail (scheduled, cancelled, waived, and other unmapped results are dropped first) and keep only the earliest one by date. If more than one event for that permit and type shares that earliest calendar day, the group counts as a fail if any of those same-day events is a fail; this is the conservative reading of a same-day tie. Later re-inspections, however many, never change that first-time outcome: a permit that passed on its first visit stays a pass even if a subsequent, unrelated inspection later fails.

Each locality's residential filter runs on that jurisdiction's own permits data, joined to its inspection log where a join is required:

• Mesa, Arizona: starts_with(permit_number,'PMT') at fetch time, then joined to dzpk-hxfb.permit_number where permit_type='RES' (unbounded pull, no date floor)
• Los Angeles, California: permit normalized (spaces to dashes) and joined to pi9x-tg5x.permit_nbr where permit_sub_type like '%Family Dwelling%' (unbounded pull, 294265 rows, no date floor needed)
• Montgomery County, Maryland: permit_type='BUILDING RESIDENTIAL PERMIT' on the inspections rows (no join needed)

The Los Angeles permits dataset used for that residential join (pi9x-tg5x, "Building Permits Issued from 2020 to Present") covers permits issued from 2020 to the present, so inspections tied to permits issued before 2020 cannot join and are not counted in the Los Angeles figures above.

Only inspection types with at least 30 first-time decisions in a locality are published; smaller cells are dropped rather than shown on a thin sample. Per-locality tables above show the 12 largest inspection types by sample size directly, with every remaining qualifying type available in an expandable table underneath, so every qualifying cell is on the page and none is silently hidden.

Administrative closure and status event types are excluded from every table above, on top of the MIN_SAMPLE gate, because they are not graded field inspections of installed work and would misrepresent what this page measures if left in:

• Los Angeles:
  • Inspection: Administrative permit-closure event, not a graded field inspection: raw rows are overwhelmingly "Permit Finaled" closure events (2 fails in 29,158 records), reviewer-verified against the live API.
  • CofO Issuance: Certificate-of-occupancy issuance/closure step, not a graded field inspection: 0 fails across 20,364 records.
  • SGSOV-Seismic Gas S/O Valve: Seismic gas shut-off valve device-registration event: 0 fails across 22,438 records, structurally inconsistent with a graded technical inspection (unlike, for example, Mesa’s "C of O" cell, which carries real fail variance and stays published).
• Montgomery County:
  • PERMIT EXPIRATION INSPECTION: Permit-status/lifecycle event checking whether a permit is nearing expiration and needs renewal, not a graded technical field inspection of installed work.

Mesa's raw inspection types were reviewed against the same pattern and none were found to be administrative closure or status events; every type published for Mesa above is a graded field check.