total_debt read LongTermDebt, else LongTermDebtNoncurrent/Current, plus one of
ShortTermBorrowings/CommercialPaper. That misses two whole tagging styles, and
it feeds net_debt -> net_debt_to_ebitda -> the categorical leverage read, so the
misses were not absences but confident wrong answers: Coca-Cola scored on 0.25bn
of commercial paper against ~39bn of debt, Verizon on 21.78bn of current
maturities against ~165bn, AT&T and Exxon produced no value at all against 134bn
and 33bn tagged. Measured over 19 large caps and 14 REITs, 11 were wrong or
absent and the rest are unchanged.
Each concept's span is now respected. LongTermDebt already includes current
maturities (Apple tags all three: 71.34 + 11.01 = 82.30), so only true
short-term borrowing is added. LongTermDebtAndCapitalLeaseObligations — what KO,
HD, T, XOM and CVX tag, and nothing read before — is noncurrent and takes a
current complement, and DebtCurrent *is* that whole complement rather than an
addition to it.
The REIT branch needed disambiguating: NotesPayable is not the same line across
issuers. MAA tags NotesPayable 5.66bn = UnsecuredDebt 5.30bn + SecuredDebt
0.36bn exactly, so there it is the total and adding the secured side
double-counts; EQR tags it alongside a larger SecuredDebt, where it is only the
unsecured component. UnsecuredDebt's presence separates them.
A component alone is no longer reported as a total. Chevron tags full debt only
in its 10-K, so its 10-Q carried 0.40bn of short-term borrowing; Boston
Properties tags SecuredDebt 4.28bn against ~15bn real. net_debt needs both sides
and yields nothing when either is missing, so None costs a leverage read where
the fragment produced a confidently wrong one.
Snapshots are immutable, so this corrects new filings only; stored history needs
scripts/reparse_fundamentals.py, which cannot complete until the EQR/931182
collision is retired.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016Lo99z3jWqu9X9ueBU3Z3D
snapshot_discrepancy named the accessions but not the columns, so it could not
distinguish "our numbers moved" from "the same filing is attributed twice".
The fields were already computed for validation_json and simply dropped from
the message; they are now in it.
A difference in cik ALONE is no longer reported as a reconstruction change at
all. Every fact matched, so two tracked CIKs are claiming one filing and the
fix is the universe, not the parser: it raises accession_cik_collision naming
both CIKs and sec_cik_overrides. It also never self-heals — the losing CIK
stores no row, so _ciks_with_snapshots never sees it and it is full-history
backfilled and re-reported every run until its ticker is re-pointed or retired.
Observed 2026-08-19 for EQR: after Equity Residential renamed to Vivmark
Residential (VMRK, CIK 906107), SEC's own company_tickers.json left the old
symbol on ERP Operating LP (CIK 931182), the non-traded co-registrant of their
combined 10-Qs. Both were tracked, both reconstructed the same two filings.
The reparse path now excludes cik-only differences from its rewrite set:
rewriting one would re-stamp the filing onto the co-registrant, taking it from
the issuer that actually filed it, which no parser fix asks for.
No stored value was wrong in that incident — reports/ carries the full
reproduction for this and for the companyfacts staleness behind the gate fix.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016Lo99z3jWqu9X9ueBU3Z3D
dennisthiessenandClaude Opus 5