Pano Coverage Audit — 75-Entity Registry

Generated: 2026-04-27 16:11 UTC Source: Pano cluster combined audit (/tmp/combined_audit.json)

Per-stage coverage at-a-glance

Coverage gaps to address

Tier 1 — substantive blockers

1. Headcount stage is undercovered (20/75 + 17/75 trend). The org_chart Gemma stage runs but emits headcount as None for the majority of entities. Likely cause: Gemma defaults to null when not confident; we accept null rather than re-prompting. Fix: add a separate Serper query layer specifically for "<entity> headcount" OR "<entity> employees" OR site:linkedin.com/company/<slug> and parse with regex.
2. Investor coverage at 68% — 24 entities have zero investors captured. Investigation: the investor stage targets organizations only, so all 12 person-entities skip it (-12). Of the remaining 63 organizations, 12 still show 0 investors. Likely public companies (SEC-only) or VC-funds that don't have visible investor stacks.

Tier 2 — minor coverage gaps

• 3 entities have <300-char backgrounds (Codametrix, Optum Bank, Optum Care). Codametrix has 10 chars — failed background generation, needs reprocessing.
• 6 entities have <4 press releases. Likely Israeli-domiciled or person-entities without press lineage.
• 7 entities have <2 SEC filings. Expected for private foreign-domiciled or sub-scale entities.

Tier 3 — observation only

• News coverage at 96% is essentially saturated; the missing 3 are person-entities with weak news presence
• Goals/RIOH coverage at 100% post-fix, validates the parser refactor

Coverage by entity-type

• Organization (63 entities): all stages populated except investors/headcount
• Person (12 entities): background, goals, RIOH all clean; investors/org/headcount stages skipped by design

Stage success ranking

1. Goals (100%) — most reliable stage post-refactor
2. RIOHs (100%) — most reliable post-fix
3. News (96%) — saturated; Serper /news endpoint works well
4. Background (96%) — 1-2 outliers
5. Press (92%) — second-tier
6. SEC (91%) — second-tier
7. Goals-clean (97%) — minor parser artifacts
8. Investors (68%) — by-design gap for persons + public companies
9. Headcount-total (27%) — major coverage gap
10. Headcount-trend (23%) — major coverage gap

Recommended action sequencing

1. Build dedicated headcount Serper layer (3-line fix in parallel_deep_research.py) — biggest coverage gain available
2. Force-refresh Codametrix background — single-entity issue with disproportionate impact on nym_medical_coding_industry_report
3. Add coverage panel to the dashboard — render this matrix at /admin/coverage so it's visible during continuous-cycle ops
4. Defer person-entity investor coverage — by-design 0%; not a real gap Co-Authored-By: Oz oz-agent@warp.dev