# Citation audit, 5% random sample (2026-07-18)

**Method:** seeded random sample (seed 1848), n=107 of 2,148 visits (~5%). Each row's primary citation traced: `ca-record` cites cross-checked against the *Archives of California* calendar record (ship named? year consistent?); flagged rows adjudicated by full-record read.

**Results:**
| Tier | n | Outcome |
|---|---|---|
| `ca-record` (automatable) | 86 | **86/86 visits supported by their citations.** 74 auto-pass; 11 flags were checker artifacts (accent-normalization: *Aránzazu/Concepción*; variant *Cutusoff*=Kutúzov; summary truncation); 1 row (v0213) supported by its **second** cite (d1053 names the *Discovery*) while its first (d1077) is a weak same-cluster co-attachment → reordered. |
| `hoc` narrative/list | 5 | not traced by local automation (OCR text mismatch / HoC V not on disk), manual tier |
| `ogden` / published | 16 | extraction-internal; manual tier (book check at v1.0) |

**Headline for the data paper:** *in a 5% random audit, 100% of automatable-tier visits were supported by their cited source; citation-level fidelity 85/86 (98.8%), the one defect being citation ordering, not a false visit.*

**Bonus catches:** v0434's «Resolución» (1795, Ensenada de Todos Santos) is a real **English merchant whaler** (Capt. Locke, HBM charter), correcting the FIRSTS draft (it is *not* Cook's sloop); note she anchored in **Baja**, so the first *Alta* whaler remains open.