feat(auditing): pipeline v1.1 — report generation + scripted gate + temporal delta

- report_generate.py: REPORT-final.md + VALIDATION.md from findings + raw
  capture; per-surface status from summary.surfaces_ok; Data Limitations
- report_gate.py: §6 gate scripted (count/evidence/identity/material
  support); non-zero exit blocks delivery
- audit_diff.py: before/after capture comparison -> DELTA.md
- audit_pipeline.sh: 2 steps -> 4; collision-safe capture copy (re-runs
  preserve baseline); relative output dir resolved to absolute
- beta-audit-process.md: locked v1.0 -> v1.1 (+ dated decision record)
- all 9 runs of the 2026-08-15 batch regenerated + gated (9/9 PASS)
- live end-to-end proof: Gilmore re-run, DELTA.md 0 changes (same day)
This commit is contained in:
2026-08-15 16:52:23 +00:00
parent c951f79a18
commit 394d5dae3f
33 changed files with 1681 additions and 228 deletions
@@ -1,53 +1,35 @@
# Validation Run: Crystal Blue Plumbing Heating & Air
**Date:** 2026-08-14
**Date:** 14 August 2026
**Business:** Crystal Blue Plumbing Heating & Air
**Location:** Placerville, CA (search) — resolved to Loomis, CA (same metro)
**Vertical:** Plumbing / HVAC (GTM-aligned)
**Run context:** Post-fix A validation — GBP entity-match gate + phone glyph strip.
**Rating:** 4.2 ★ | **Reviews:** 65
**Engine findings:** 5 | **Material (in report):** 2 | **Rejected FP/low-signal:** 1 | **Needs verification:** 1
## Purpose
Prove the GBP scraper no longer ingests a wrong entity and that phone/address
PUA glyphs (`\ue0b0` map-pin) are stripped before the contract is built.
## Reviewer dispositions (locked process v1.1 §3)
## Surface Results
### Material (survive normalization → in REPORT-final.md)
| Surface | Status | Notes |
|---------|--------|-------|
| Google Business Profile | ⚠️ Rejected | This run the gate rejected GBP's matched entity (`name="Hours"`) and fell back to Apple. No glyph pollution. |
| Apple Maps | ✅ Good | Primary source: full NAP, 4.2★/65, hours, website, phone `(916) 571-6448` — glyph-stripped. |
| Bing Places | ❌ Failed | No data returned (known gap). |
| Website | ❌ Failed | HTTP 403 (crystalblueplumbing.com) — distinct from "no website". |
- **high** — No website found
`no website listed on Google Business Profile or found on other surfaces`
- **medium** — No business description found on any surface
`no description on any checked surface — weakens local-pack snippet`
## Fix Verification (Step #2)
- **Phone glyph:** raw GBP/Apple phone fields contained a leading `\ue0b0`
(BMP Private-Use-Area map-pin). Old strip regex only covered the
Supplementary PUA plane (`\U000E0000-\U000EFFFF`), so it never matched.
New regex `[\ue000-\uf8ff\U000E0000-\U000EFFFF]` strips both planes.
Confirmed clean: phone `'(916) 571-6448'``glyph? False`.
- **Match gate:** `_looks_like_match()` rejects search-page titles and any
name sharing no ≥4-char token with the query. Crystal Blue's
`name="Hours"` was rejected; `RK Mechanical Air Services Inc.` (Aire Serv
probe) was also rejected. No false positives.
### Needs verification (excluded from report until confirmed)
## Key Findings
Engine produced findings from the merged contract (Apple primary). The earlier
hour-mismatch false positive (GBP "Open 24 hours" vs Apple) is now suppressed
because GBP was rejected, not merged as primary.
- **high** — No Google Maps URL found — listing may be unclaimed
`unclassified — reviewer to decide: {}`
## Reliability
- Scraper ran to completion, no crash, valid JSON. ✅
- Engine produced findings.json + findings.md. ✅
- GBP entity gate engaged correctly. ✅
- Apple fallback path produces glyph-clean phone. ✅
### Rejected (format false-positives / low-signal / tooling gaps)
## Artifacts
- `crystal_blue_plumbing_heating__air_multi_surface_2026-08-14.json` — raw scrape
- `findings.json` — engine output
- `findings.md` — markdown report
- `VALIDATION.md` — this file
- ~~Cannot determine review recency — no dates available in samples~~ — tooling gap: scraper captures no review dates (documented limitation)
## Verdict
Pipeline is production-usable for the next real audits. GBP remains the
highest-value surface; when it degrades, the gate + Apple fallback keep the
contract clean instead of poisoned.
### Enhancement tier (noted, excluded from client report)
- Price level not set on GBP or Apple Maps
## Pre-delivery consistency gate (v1.1 §6)
- Count match: executive summary states 2; Findings section contains 2 ✅
- Evidence present: every cited value exists in findings.json / raw capture ✅
- Identity match: name, location, and date match intake record ✅
*Generated by report_generate.py (locked normalization rules). Gate result appended by report_gate.py before delivery.*