fix(sec): correct fundamentals derivation from SEC company facts

The A5 parity report surfaced coverage gaps and wrong values that all traced
to the SEC facts parser and read-time derivation rather than to bad source
data. Fixes, each validated by replaying the production parser + derivation
against live company facts:

- Period identity is derived from period_end against the issuer's fiscal
  calendar, not SEC's fy/fp fields, which collide (two period ends on one key,
  one silently discarded) and invert (a period sorting before one that precedes
  it) often enough to break the quarter chain. Recovers BXP, CRM, CRWD, FRT,
  MTD, NTAP, PPL, STX, WDAY. Fixed labels are internal ordering keys only (not
  in any API schema), so a filer whose year ends in early January shifting by
  one is harmless.
- Revenue concept list gains RevenuesNetOfInterestExpense (banks) and the
  IncludingAssessedTax variant (REITs/consumer); EPS gains the continuing-ops
  variant (REG/FCX) and, last, basic EPS for a period tagging no diluted
  variant at all (PPL). All appended, so any issuer that already resolved keeps
  its concept.
- YTD span tolerance 20 -> 25 days, covering 4-4-5 retail calendars whose
  36-week YTD-Q3 (251-252d) previously missed by ~2 (COST, PEP, DPZ).
- Amendment resolution is per field: a partial 10-K/A (Part III only, no
  financial facts) no longer blanks the period (DVN).
- TTM diluted EPS is suppressed when a split contaminates the trailing window
  (BKNG's mixed-unit sum produced a P/E of 1.10 that clamped to a perfect
  fundamental sub-score). A post-filing split with no share-count evidence
  (KLAC) remains undetectable from this data.
- Multi-class share fallback: weighted_avg_diluted_shares is captured and used
  for market cap when the cover-page count is absent (dimensional, so missing
  from company facts for META/CMCSA/CHTR/FOXA/NWSA/LEN). Within ~0.6% of the
  true count on controls; flagged shares_estimated in the API. BRK-B has no
  weighted-average fact either and stays unavailable.

820 unit tests pass; new tests confirmed to fail against the pre-fix code.
Effect is inert until existing rows are reparsed (see reparse path).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

alembic/versions/027_snapshot_weighted_avg_diluted_shares.py

@@ -0,0 +1,43 @@
+"""fundamental_snapshots.weighted_avg_diluted_shares — market-cap fallback
+
+Revision ID: 027
+Revises: 026
+Create Date: 2026-07-24 00:00:00.000000
+
+Multi-class issuers report the cover-page share count per share class. That is a
+dimensional fact and Company Facts is non-dimensional, so it is absent entirely:
+META has never tagged it, CMCSA stops in 2009, BRK-B in 2011, CHTR in 2016 (when
+the Time Warner Cable deal made it multi-class). `shares_outstanding` is
+therefore null for a large slice of the mega-cap universe, which silently removes
+both `market_cap_est` and `fcf_yield`.
+
+The weighted-average diluted count is always present (EPS requires it) and is
+consolidated across classes. Measured against issuers where the true
+point-in-time count IS available, it lands within ~0.6%: GOOGL 0.9936, MRNA
+1.0045, AAPL 0.9974, MSFT 0.9978.
+
+Stored as its own column rather than backfilled into `shares_outstanding`, so the
+point-in-time column keeps its strict meaning and the fallback stays an explicit,
+labelled read-time decision. Existing rows are null until a reparse.
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision: str = "027"
+down_revision: Union[str, None] = "026"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "fundamental_snapshots",
+        sa.Column("weighted_avg_diluted_shares", sa.Float(), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("fundamental_snapshots", "weighted_avg_diluted_shares")

app/models/fundamental_snapshot.py

@@ -12,8 +12,10 @@ class FundamentalSnapshot(Base):
     Keyed by issuer (CIK), not ticker — multi-class issuers (GOOG/GOOGL) share
     one CIK and one set of fundamentals; the ``tickers.cik`` column is the only
     join point. Amendments are retained: every accession is a distinct immutable
-    row, and readers pick the newest valid ``accepted_at`` per
+    row, and readers resolve (cik, fiscal_year, fiscal_period) at read time by
-    (cik, fiscal_year, fiscal_period) at read time — no flags, no mutation.
+    taking the newest ``accepted_at`` **per field**, falling back to the newest
+    accession that actually reports one — a partial amendment (a 10-K/A adding
+    Part III reports no financial facts) must not blank the period — no flags, no mutation.
 
     **Facts are stored as the filing reports them, never as derived quarters.**
     Duration facts (revenue, net_income, operating_income, diluted_eps, cfo,
@@ -70,6 +72,12 @@ class FundamentalSnapshot(Base):
     # reported "as of" its own date, which can differ from period_end — store it
     # so market cap uses the right point-in-time count.
     shares_outstanding_date: Mapped[date | None] = mapped_column(Date, nullable=True)
+    # Weighted-average diluted count for the filing's most recent quarter — the
+    # market-cap fallback when the cover-page count is absent, which it always is
+    # for multi-class issuers (per-class facts are dimensional, and companyfacts
+    # is not). An average is not cumulative, so unlike the duration facts above
+    # this is NOT a YTD value: it is the shortest-span fact ending at period_end.
+    weighted_avg_diluted_shares: Mapped[float | None] = mapped_column(Float, nullable=True)
 
     import_run_id: Mapped[int | None] = mapped_column(
         ForeignKey("data_import_runs.id", ondelete="SET NULL"), nullable=True

app/services/fundamentals_api_service.py

@@ -155,6 +155,13 @@ def _build_valuation(derived, subject_price, peer_derived, peer_price_by_cik, tw
         "pe": _round(pe, 2),
         "fcf_yield": _round(fcf_yield, 2),
         "market_cap_est": _round(market_cap, 0),
+        # market_cap_est and fcf_yield both rest on the share count. When it came
+        # from the weighted-average diluted fallback (multi-class issuers, whose
+        # per-class cover-page count is absent from companyfacts), say so rather
+        # than presenting a period average as a point-in-time count.
+        "shares_estimated": bool(
+            market_cap is not None and derived.shares_outstanding_estimated
+        ),
         "pe_industry": pe_industry,
         "fcf_yield_industry": fcf_yield_industry,
         "price_date": _iso(price_date),

app/services/fundamentals_derivation.py

@@ -8,8 +8,10 @@ schema decision. No I/O, no DB: it takes an issuer's snapshot rows (ORM rows or
 any objects with the same attributes) and returns structured metrics.
 
 Rules:
-- **Amendment selection:** for each (fiscal_year, fiscal_period), the row with
+- **Amendment selection:** for each (fiscal_year, fiscal_period), the newest
-  the newest `accepted_at` wins.
+  `accepted_at` wins **per field**, falling back to the newest row that actually
+  reports one. A partial amendment (a 10-K/A adding Part III carries no financial
+  facts) must not blank the period.
 - **Discrete quarter** = YTD(Qn) − YTD(Qn−1); Q1 = YTD(Q1); **Q4 = YTD(FY) −
   YTD(Q3)**. Any missing period → the derived value is null, never partial.
 - **TTM** = sum of the trailing four discrete quarters ending at a period.
@@ -21,6 +23,7 @@ from __future__ import annotations
 
 from dataclasses import dataclass, field
 from datetime import date
+from types import SimpleNamespace
 from typing import Any, Iterable
 
 _FP_TO_Q = {"Q1": 1, "Q2": 2, "Q3": 3, "FY": 4}
@@ -38,6 +41,20 @@ _FLOW_FIELDS = (
     "revenue", "net_income", "operating_income", "diluted_eps", "cfo", "capex",
     "depreciation_amortization",
 )
+# Reported facts resolved independently across a period's accessions (see
+# _merge_amendments); period identity/provenance is taken from the newest one.
+_MERGED_FIELDS = (
+    *_FLOW_FIELDS,
+    "cash_and_st_investments", "total_debt", "shares_outstanding",
+    "shares_outstanding_date",
+    # period_start is set alongside revenue by the parser, so it follows the same
+    # fallback: a bare amendment reports neither and must not blank it.
+    "period_start",
+)
+_CARRIED_FIELDS = (
+    "fiscal_year", "fiscal_period", "period_end", "filed_date",
+    "accepted_at", "form", "accession", "cik",
+)
 
 
 @dataclass
@@ -60,8 +77,15 @@ class DerivedFundamentals:
     metrics: dict[str, MetricSeries] = field(default_factory=dict)
     # request-time valuation inputs (ratios are computed in the API with price)
     ttm_diluted_eps: float | None = None
+    # Set when ttm_diluted_eps was suppressed rather than simply unavailable.
+    ttm_diluted_eps_caveat: str | None = None
     ttm_fcf: float | None = None
     shares_outstanding: float | None = None
+    # True when shares_outstanding came from the weighted-average diluted count
+    # because the point-in-time cover-page count was absent (always so for
+    # multi-class issuers). Consumers must label anything derived from it as
+    # estimated — it is a period average, not a point-in-time count.
+    shares_outstanding_estimated: bool = False
     latest_period_end: date | None = None
     latest_filed_date: date | None = None
 
@@ -85,6 +109,15 @@ def derive(snapshots: Iterable[Any]) -> DerivedFundamentals:
     result.latest_period_end = latest_row.period_end
     result.latest_filed_date = latest_row.filed_date
     result.shares_outstanding = getattr(latest_row, "shares_outstanding", None)
+    if result.shares_outstanding is None:
+        # Multi-class issuers (META, CMCSA, BRK-B, CHTR, FOXA, NWSA, LEN) report
+        # the cover-page count per class, which is dimensional and so absent from
+        # companyfacts — leaving market cap and FCF yield silently unavailable for
+        # some of the largest names. The weighted-average diluted count is always
+        # present and within ~0.6% of the true count where both exist, so fall
+        # back to it and mark the result estimated rather than show nothing.
+        result.shares_outstanding = getattr(latest_row, "weighted_avg_diluted_shares", None)
+        result.shares_outstanding_estimated = result.shares_outstanding is not None
     result.ttm_diluted_eps = _ttm(discrete["diluted_eps"], *latest)
     ttm_cfo = _ttm(discrete["cfo"], *latest)
     ttm_capex = _ttm(discrete["capex"], *latest)
@@ -102,7 +135,14 @@ def derive(snapshots: Iterable[Any]) -> DerivedFundamentals:
         "net_debt_to_ebitda": _leverage_series(selected, discrete, tape),
         "share_count_change_yoy": _share_change_series(selected, tape),
     }
-    _guard_split_sensitive_metrics(result.metrics)
+    # TTM EPS sums four quarters of *per-share* values, so a split inside that
+    # window mixes pre- and post-split units — the same distortion the guard
+    # already catches for the series, and the one that produced BKNG's P/E of
+    # 1.10. Left unguarded it does not merely mislead: a nonsense-low P/E clamps
+    # to a perfect 100 fundamental sub-score, so it must null out like the rest.
+    if _guard_split_sensitive_metrics(result.metrics):
+        result.ttm_diluted_eps = None
+        result.ttm_diluted_eps_caveat = SPLIT_SENSITIVE_CAVEAT
     for series in result.metrics.values():
         series.period_end = latest_row.period_end
         series.filed_date = latest_row.filed_date
@@ -112,17 +152,57 @@ def derive(snapshots: Iterable[Any]) -> DerivedFundamentals:
 # -- period selection --------------------------------------------------------
 
 def _select_latest_per_period(snapshots: Iterable[Any]) -> dict[tuple[int, str], Any]:
-    best: dict[tuple[int, str], Any] = {}
+    grouped: dict[tuple[int, str], list[Any]] = {}
     for row in snapshots:
         fp = getattr(row, "fiscal_period", None)
         fy = getattr(row, "fiscal_year", None)
         if fp not in _FP_TO_Q or fy is None:
             continue
-        key = (fy, fp)
+        grouped.setdefault((fy, fp), []).append(row)
-        cur = best.get(key)
+    return {key: _merge_amendments(rows) for key, rows in grouped.items()}
-        if cur is None or _accepted(row) > _accepted(cur):
+
-            best[key] = row
+
-    return best
+def _merge_amendments(rows: list[Any]) -> Any:
+    """Resolve one period from its accessions: newest wins, per field.
+
+    Amendments are frequently partial — a 10-K/A filed only to add Part III
+    reports no financial facts at all. Taking the newest accession wholesale
+    would blank every field it omits and null the period downstream (and with
+    it TTM and YoY, which need an unbroken quarter chain), so each field falls
+    back to the newest accession that actually reports it.
+
+    Only rows sharing the newest row's ``period_end`` are merged. A same-key row
+    covering a *different* period is a mislabelled filing, not an amendment, and
+    blending the two would silently mix fiscal years.
+    """
+    if len(rows) == 1:
+        return rows[0]
+    ordered = sorted(rows, key=_amendment_order, reverse=True)  # newest first
+    newest = ordered[0]
+    same_period = [
+        row
+        for row in ordered
+        if getattr(row, "period_end", None) == getattr(newest, "period_end", None)
+    ]
+    if len(same_period) == 1:
+        return newest
+    merged = SimpleNamespace(**{name: getattr(newest, name, None) for name in _CARRIED_FIELDS})
+    for name in _MERGED_FIELDS:
+        merged_value = None
+        for row in same_period:  # newest first
+            value = getattr(row, name, None)
+            if value is not None:
+                merged_value = value
+                break
+        setattr(merged, name, merged_value)
+    return merged
+
+
+def _amendment_order(row: Any) -> tuple[bool, Any]:
+    # (has-timestamp, timestamp) so a row without one sorts oldest instead of
+    # raising when compared against a row that has one.
+    accepted = _accepted(row)
+    return (accepted is not None, accepted)
 
 
 def _accepted(row: Any):
@@ -255,18 +335,21 @@ def _share_change_series(selected, tape) -> MetricSeries:
     return _series(pts)
 
 
-def _guard_split_sensitive_metrics(metrics: dict[str, MetricSeries]) -> None:
+def _guard_split_sensitive_metrics(metrics: dict[str, MetricSeries]) -> bool:
     """Suppress historical comparisons likely distorted by a corporate action.
 
     Company Facts has no point-in-time split factors. A large YoY share-count
     move can therefore make both the point-in-time share comparison and
     per-share EPS growth non-comparable. Keep the raw facts in snapshots, but
     expose nulls plus an explicit caveat in the user-facing derived series.
+
+    Returns True when the *latest* period is suspect, so callers can apply the
+    same suppression to per-share scalars derived from that window.
     """
     shares = metrics.get("share_count_change_yoy")
     eps = metrics.get("eps_growth_yoy")
     if shares is None or eps is None:
-        return
+        return False
 
     suspect_periods = {
         point.period_end
@@ -275,18 +358,21 @@ def _guard_split_sensitive_metrics(metrics: dict[str, MetricSeries]) -> None:
         and abs(point.value) >= SPLIT_SUSPECT_SHARE_CHANGE_PCT
     }
     if not suspect_periods:
-        return
+        return False
 
+    latest_suspect = False
     for series in (shares, eps):
         latest_guarded = bool(
             series.history and series.history[-1].period_end in suspect_periods
         )
+        latest_suspect = latest_suspect or latest_guarded
         for point in series.history:
             if point.period_end in suspect_periods:
                 point.value = None
         series.value = series.history[-1].value if series.history else None
         if latest_guarded:
             series.caveat = SPLIT_SENSITIVE_CAVEAT
+    return latest_suspect
 
 
 def _net_debt(row: Any) -> float | None:

app/services/sec_facts_parser.py

@@ -8,6 +8,9 @@ fixture and verifiable against a real companyfacts pull.
 The load-bearing rules (design Decision 2 + review):
 - Period identity comes from `end == submissions.reportDate`, never `fy/fp`
   (fy/fp is the *filing's* context; comparatives inside a filing repeat it).
+  This applies to the stored `fiscal_year`/`fiscal_period` too: they are derived
+  from `reportDate` against the issuer's `fiscalYearEnd` (see `_period_identity`),
+  because SEC's fy/fp collide and invert often enough to break the quarter chain.
 - Duration facts are stored as **cumulative YTD**: pick the fact whose span
   matches the fiscal-period-to-date length (Q1≈3mo … FY≈12mo) within tolerance.
   If no YTD-length fact exists, store null — never a discrete masquerading as YTD.
@@ -15,7 +18,10 @@ The load-bearing rules (design Decision 2 + review):
   is a single consolidated value: the cover-page `dei` fact (its own cover-date
   `end` stored separately) if present, else `us-gaap:CommonStockSharesOutstanding`
   at period end (e.g. Alphabet has no `dei` fact) — never a class sum or the
-  weighted-average/diluted count.
+  weighted-average/diluted count. Multi-class issuers report it per class, which
+  is dimensional and therefore absent from companyfacts entirely, so
+  `weighted_avg_diluted_shares` is stored alongside as an explicit fallback for
+  market cap — a separate column, never backfilled into `shares_outstanding`.
 - Cash and debt composites are aggregate-first and mutually exclusive (each
   source tag counted at most once).
 
@@ -29,7 +35,7 @@ from __future__ import annotations
 import logging
 import math
 from dataclasses import dataclass, field
-from datetime import date, datetime
+from datetime import date, datetime, timedelta
 from typing import Any, NamedTuple
 
 logger = logging.getLogger(__name__)
@@ -37,14 +43,35 @@ logger = logging.getLogger(__name__)
 # Expected YTD span (days) per fiscal period; a duration fact must land within
 # tolerance of this to count as the period's cumulative value.
 _EXPECTED_YTD_DAYS = {"Q1": 91, "Q2": 182, "Q3": 273, "FY": 365}
-_YTD_TOLERANCE_DAYS = 20  # covers 52/53-week fiscal calendars
+# Period identity (see _period_identity): how far a quarter end sits before its
+# fiscal-year end, and how far a fiscal-year end may drift from the nominal MMDD.
+# The quarter bands are 91 days apart, so ±35 stays unambiguous even for a 4-4-5
+# filer whose 16-week Q4 puts Q3 112 days out.
+_QUARTER_DAYS_TO_FY_END = {"Q1": 273, "Q2": 182, "Q3": 91}
+_QUARTER_TOLERANCE_DAYS = 35
+_FYE_DRIFT_TOLERANCE_DAYS = 21
+# Covers 52/53-week calendars *and* 4-4-5 retail ones (12/12/12/16 weeks), whose
+# YTD-Q3 is 36 weeks = 251-252 days and missed a 20-day tolerance by ~2 -- so
+# COST/PEP lost Q3 every year, breaking the quarter chain and nulling TTM + YoY.
+# Q1 84d, Q2 168d and FY 364d were always inside. Adjacent periods stay
+# unambiguous at 25 (66-116, 157-207, 248-298, 340-390).
+_YTD_TOLERANCE_DAYS = 25
 
 # us-gaap duration concepts (money), priority order; first present wins.
 _DURATION_USD = {
+    # Order is load-bearing (first present wins) and the tail entries are
+    # deliberately *appended*: every issuer that already resolved keeps the same
+    # concept, and only issuers that resolved to nothing gain a value.
+    #   - IncludingAssessedTax: REITs/consumer filers that tag only this variant
+    #     (e.g. ARE, KHC) reported no revenue at all.
+    #   - RevenuesNetOfInterestExpense: the banks' total-revenue tag. JPM/GS/WFC
+    #     tag it in every 10-Q and `Revenues` only (if at all) in the 10-K.
     "revenue": [
         "RevenueFromContractWithCustomerExcludingAssessedTax",
         "Revenues",
         "SalesRevenueNet",
+        "RevenueFromContractWithCustomerIncludingAssessedTax",
+        "RevenuesNetOfInterestExpense",
     ],
     "net_income": ["NetIncomeLoss"],
     "operating_income": ["OperatingIncomeLoss"],
@@ -62,7 +89,27 @@ _DURATION_USD = {
         "DepreciationAndAmortization",
     ],
 }
-_EPS_CONCEPTS = ["EarningsPerShareDiluted"]  # unit USD/shares
+# unit USD/shares. Appended (not reordered) so any issuer that already resolved
+# keeps the same concept. REG tags only the continuing-operations variant on every
+# filing; FCX switches by form type -- EarningsPerShareDiluted in its 10-Qs, the
+# continuing-ops tag in its 10-K -- which nulled the FY row and killed Q4 + TTM.
+# The basic variants are a last resort for a period that tags no diluted EPS at
+# all (PPL's 2026 Q1). Basic ignores option/convert dilution so it slightly
+# overstates EPS (~1.2% for PPL), but only fires when diluted is entirely absent,
+# and high-dilution names always tag diluted -- so it never displaces a real one.
+_EPS_CONCEPTS = [
+    "EarningsPerShareDiluted",
+    "IncomeLossFromContinuingOperationsPerDilutedShare",
+    "EarningsPerShareBasic",
+    "IncomeLossFromContinuingOperationsPerBasicShare",
+]
+# Weighted-average diluted share count (unit "shares"), the market-cap fallback
+# for multi-class issuers whose cover-page count is dimensional and therefore
+# absent from companyfacts. Always present, since EPS is computed from it.
+_WEIGHTED_AVG_SHARE_CONCEPTS = [
+    "WeightedAverageNumberOfDilutedSharesOutstanding",
+    "WeightedAverageNumberOfSharesOutstandingBasicAndDiluted",
+]
 # us-gaap instant (balance-sheet) concepts, at end == reportDate.
 _CASH = ["CashAndCashEquivalentsAtCarryingValue"]
 _ST_INVESTMENTS = ["ShortTermInvestments", "MarketableSecuritiesCurrent"]  # pick one
@@ -104,6 +151,7 @@ class SnapshotRow:
     total_debt: float | None = None
     shares_outstanding: float | None = None
     shares_outstanding_date: date | None = None
+    weighted_avg_diluted_shares: float | None = None
 
 
 @dataclass
@@ -127,9 +175,14 @@ def parse_snapshots(
     companyfacts: dict[str, Any],
     filings: dict[str, FilingMeta],
     accessions: set[str],
+    fiscal_year_end: str | None = None,
 ) -> ParseResult:
     """Build snapshot rows for ``accessions`` (those with facts + filing meta).
 
+    ``fiscal_year_end`` is the issuer's ``submissions.fiscalYearEnd`` (MMDD) and
+    is what makes period identity independent of SEC's unreliable fy/fp fields
+    (see ``_period_identity``). Omitting it falls back to the old fy/fp behaviour.
+
     ``skipped_filings`` = no row produced (missing facts/meta or no usable period
     identity); ``field_issues`` = a row was produced but a field is null/ambiguous.
     Callers must not use field issues as failed-row coverage.
@@ -143,7 +196,7 @@ def parse_snapshots(
         if meta is None or not facts:
             result.skipped_filings.append({"accession": accn, "reason": "no facts or filing metadata"})
             continue
-        row, note = _parse_one(cik, accn, facts, meta)
+        row, note = _parse_one(cik, accn, facts, meta, fiscal_year_end)
         if row is None:
             result.skipped_filings.append({"accession": accn, "reason": note or "unparseable"})
             continue
@@ -190,12 +243,18 @@ def _index_by_accession(companyfacts: dict[str, Any]) -> dict[str, list[Fact]]:
 
 
 def _parse_one(
-    cik: str, accn: str, facts: list[Fact], meta: FilingMeta
+    cik: str, accn: str, facts: list[Fact], meta: FilingMeta,
+    fiscal_year_end: str | None = None,
 ) -> tuple[SnapshotRow | None, str | None]:
     """Returns (row, note). row is None when there's no usable period identity;
     note is a validation reason (row-skip reason when row is None, else a
     field-level issue such as ambiguous shares)."""
-    fy, fp = _fiscal_context(facts, meta.report_date)
+    fy, fp = _period_identity(meta, fiscal_year_end)
+    if fy is None or fp is None:
+        # No fiscal calendar, or a period the calendar cannot place (a transition
+        # period). Fall back to the filing's own context: an imperfect label still
+        # beats dropping the filing entirely.
+        fy, fp = _fiscal_context(facts, meta.report_date)
     if fy is None or fp not in _EXPECTED_YTD_DAYS:
         return None, "no usable period identity"
 
@@ -227,9 +286,80 @@ def _parse_one(
     shares, shares_date, ambiguous = _select_shares(facts, meta.report_date)
     row.shares_outstanding = shares
     row.shares_outstanding_date = shares_date
+    row.weighted_avg_diluted_shares = _select_weighted_avg_shares(facts, meta.report_date)
+
     return row, ("ambiguous shares outstanding" if ambiguous else None)
 
 
+def _period_identity(
+    meta: FilingMeta, fiscal_year_end: str | None
+) -> tuple[int | None, str | None]:
+    """(fiscal_year, fiscal_period) from the period end and the issuer's fiscal
+    calendar — never from the fy/fp fields.
+
+    SEC's fy/fp describe the *filing*, and they are unreliable as period identity:
+    observed in production, a 10-Q labelled ``FY`` (BXP), a year ending 2025-12-31
+    labelled 2024 (FRT, a December filer), a year ending 2025-06-27 labelled 2027
+    (STX), and four different period ends all labelled 2022 Q3 (PPL). Because
+    readers key on (fiscal_year, fiscal_period), colliding labels silently discard
+    a period and inverted ones scramble the quarter chain — nulling TTM and YoY.
+
+    ``period_end`` is authoritative, so identity is derived from it: the form
+    decides FY vs quarter, and distance to the fiscal-year end decides which
+    quarter. Labels need not match the issuer's own naming — a filer whose year
+    ends in early January (DPZ) shifts by one — they need to be unique, monotonic
+    and YoY-aligned, which is all the derivation asks of them. Nothing outside the
+    derivation reads these columns.
+    """
+    fy = _fiscal_year_of(meta.report_date, fiscal_year_end)
+    if fy is None:
+        return None, None
+    if meta.form.startswith("10-K"):
+        return fy, "FY"
+    nominal_end = _nominal_fy_end(fy, fiscal_year_end)
+    if nominal_end is None:
+        return None, None
+    remaining = (nominal_end - meta.report_date).days
+    best = min(
+        _QUARTER_DAYS_TO_FY_END,
+        key=lambda k: abs(_QUARTER_DAYS_TO_FY_END[k] - remaining),
+    )
+    if abs(_QUARTER_DAYS_TO_FY_END[best] - remaining) > _QUARTER_TOLERANCE_DAYS:
+        return None, None  # transition period or odd filing — let the caller fall back
+    return fy, best
+
+
+def _nominal_fy_end(year: int, fiscal_year_end: str | None) -> date | None:
+    """The issuer's nominal fiscal-year end in ``year`` from a MMDD string."""
+    if not fiscal_year_end or len(fiscal_year_end) != 4 or not fiscal_year_end.isdigit():
+        return None
+    month, day = int(fiscal_year_end[:2]), int(fiscal_year_end[2:])
+    if not 1 <= month <= 12 or not 1 <= day <= 31:
+        return None
+    while day > 28:  # 52/53-week ends land on 0229/0230/0231 in some filings
+        try:
+            return date(year, month, day)
+        except ValueError:
+            day -= 1
+    return date(year, month, day)
+
+
+def _fiscal_year_of(period_end: date, fiscal_year_end: str | None) -> int | None:
+    """Which fiscal year ``period_end`` belongs to.
+
+    A 52/53-week calendar's real year end drifts around the nominal MMDD (and can
+    cross the calendar year), so allow drift before rolling into the next year.
+    """
+    nominal = _nominal_fy_end(period_end.year, fiscal_year_end)
+    if nominal is None:
+        return None
+    return (
+        period_end.year
+        if period_end <= nominal + timedelta(days=_FYE_DRIFT_TOLERANCE_DAYS)
+        else period_end.year + 1
+    )
+
+
 def _fiscal_context(facts: list[Fact], report_date: date) -> tuple[int | None, str | None]:
     """The filing's (fy, fp) taken as the majority context among the facts that
     end at reportDate (the current-period facts, which share the filing's
@@ -352,6 +482,34 @@ def _select_shares(
     return None, None, False  # simply absent — not a conflict
 
 
+def _select_weighted_avg_shares(facts: list[Fact], report_date: date) -> float | None:
+    """The most recent quarter's weighted-average diluted share count.
+
+    Deliberately the **shortest** duration ending at reportDate, not the YTD one:
+    the shorter the window the closer the average sits to the current count, which
+    is what a market cap wants. Measured against issuers where the true
+    point-in-time count is available, the quarter average is within ~0.6%.
+    """
+    best: tuple[int, float] | None = None
+    for concept in _WEIGHTED_AVG_SHARE_CONCEPTS:
+        for f in facts:
+            if (
+                f.taxonomy != "us-gaap"
+                or f.concept != concept
+                or f.unit != "shares"
+                or f.start is None
+                or f.end != report_date
+                or f.val <= 0
+            ):
+                continue
+            span = (f.end - f.start).days
+            if best is None or span < best[0]:
+                best = (span, float(f.val))
+        if best is not None:
+            return best[1]  # first present concept wins, as elsewhere
+    return None
+
+
 def _d(value: Any) -> date | None:
     if not value:
         return None

tests/unit/test_fundamentals_derivation.py

@@ -29,6 +29,7 @@ class Snap:
     cash_and_st_investments: float | None = None
     total_debt: float | None = None
     shares_outstanding: float | None = None
+    weighted_avg_diluted_shares: float | None = None
 
 
 _FP = ["Q1", "Q2", "Q3", "FY"]
@@ -188,3 +189,112 @@ def test_amendment_selection_newest_accepted_wins():
     # Q4 revenue discrete now uses the amended YTD(FY)=999999 minus YTD(Q3)=363
     # so TTM/growth reflects the amendment, proving newest accepted_at won.
     assert d.metrics["revenue_growth_yoy"].value != pytest.approx(10.0, abs=1e-6)
+
+
+# -- partial amendments (A5 parity findings) ---------------------------------
+
+def test_partial_amendment_does_not_blank_the_period():
+    # DVN's FY2025 10-K/A carries no financial facts at the report date. Taking
+    # the newest accession wholesale nulled the period, and with it the quarter
+    # chain, TTM and YoY.
+    rows = _two_years()
+    part_iii_only = Snap(2026, "FY", date(2026, 9, 30), date(2026, 11, 1),
+                         datetime(2027, 1, 1, tzinfo=UTC))
+    baseline = fd.derive(rows)
+    d = fd.derive(rows + [part_iii_only])
+    assert d.ttm_diluted_eps == pytest.approx(baseline.ttm_diluted_eps)
+    assert d.metrics["revenue_growth_yoy"].value == pytest.approx(
+        baseline.metrics["revenue_growth_yoy"].value
+    )
+
+
+def test_amendment_restating_one_field_leaves_the_others_intact():
+    rows = _two_years()
+    revenue_only = Snap(2026, "FY", date(2026, 9, 30), date(2026, 11, 1),
+                        datetime(2027, 1, 1, tzinfo=UTC), revenue=999999)
+    baseline = fd.derive(rows)
+    d = fd.derive(rows + [revenue_only])
+    assert d.metrics["revenue_growth_yoy"].value != pytest.approx(
+        baseline.metrics["revenue_growth_yoy"].value
+    )
+    assert d.ttm_diluted_eps == pytest.approx(baseline.ttm_diluted_eps)  # fell back
+
+
+def test_same_key_row_for_a_different_period_is_never_merged():
+    # SEC labels two different year-ends with one fiscal_year for some filers
+    # (FRT, CRM). That is a mislabelled filing, not an amendment -- merging the
+    # two would silently blend fiscal years.
+    rows = _two_years()
+    mislabelled = Snap(2026, "FY", date(2027, 9, 30), date(2027, 11, 1),
+                       datetime(2027, 12, 1, tzinfo=UTC), revenue=999999)
+    selected = fd._select_latest_per_period(rows + [mislabelled])
+    assert selected[(2026, "FY")] is mislabelled
+
+
+# -- split safety for the TTM EPS scalar (A5 parity findings) ----------------
+
+def _split_rows():
+    """Two years where the share count jumps ~25x at the latest quarter, as
+    BKNG's did (31.7M -> 774.9M) when its split landed mid-window."""
+    rows = _two_years()
+    for row in rows:
+        if (row.fiscal_year, row.fiscal_period) == (2026, "FY"):
+            row.shares_outstanding = 25000.0  # vs 1000 a year earlier
+    return rows
+
+
+def test_split_suppresses_ttm_diluted_eps():
+    # TTM sums four quarters of per-share values; a split inside the window
+    # mixes units. Unguarded this produced BKNG's P/E of 1.10, which clamps to a
+    # *perfect* fundamental sub-score -- worse than having no value at all.
+    d = fd.derive(_split_rows())
+    assert d.ttm_diluted_eps is None
+    assert d.ttm_diluted_eps_caveat == fd.SPLIT_SENSITIVE_CAVEAT
+
+
+def test_ttm_diluted_eps_survives_when_no_split_is_suspected():
+    d = fd.derive(_two_years())
+    assert d.ttm_diluted_eps is not None
+    assert d.ttm_diluted_eps_caveat is None
+
+
+def test_split_guard_leaves_dollar_scalars_alone():
+    # Only per-share values are split-sensitive; FCF is in dollars.
+    baseline = fd.derive(_two_years())
+    d = fd.derive(_split_rows())
+    assert d.ttm_fcf == pytest.approx(baseline.ttm_fcf)
+
+
+# -- multi-class share-count fallback (A5 parity findings) -------------------
+
+def test_shares_fall_back_to_weighted_average_when_cover_page_count_is_absent():
+    # META/CMCSA/BRK-B/CHTR report the cover-page count per share class, which is
+    # dimensional and therefore absent from companyfacts -- silently removing
+    # market cap and FCF yield for some of the largest issuers.
+    rows = _two_years()
+    for row in rows:
+        row.shares_outstanding = None
+        row.weighted_avg_diluted_shares = 2_564_000_000.0
+    d = fd.derive(rows)
+    assert d.shares_outstanding == 2_564_000_000.0
+    assert d.shares_outstanding_estimated is True
+
+
+def test_point_in_time_share_count_is_preferred_and_not_flagged():
+    baseline = fd.derive(_two_years()).shares_outstanding
+    assert baseline is not None, "fixture should carry a cover-page count"
+    rows = _two_years()
+    for row in rows:
+        row.weighted_avg_diluted_shares = 1.0  # must lose to the real count
+    d = fd.derive(rows)
+    assert d.shares_outstanding == baseline
+    assert d.shares_outstanding_estimated is False
+
+
+def test_no_share_count_at_all_stays_none_and_unflagged():
+    rows = _two_years()
+    for row in rows:
+        row.shares_outstanding = None
+    d = fd.derive(rows)
+    assert d.shares_outstanding is None
+    assert d.shares_outstanding_estimated is False

tests/unit/test_sec_facts_parser.py

@@ -270,3 +270,196 @@ async def test_live_apple_parse_invariants():
     # shares cover-date differs from period_end
     latest = max(rows, key=lambda r: r.period_end)
     assert latest.shares_outstanding_date != latest.period_end
+
+
+# -- revenue concept coverage (A5 parity findings) ---------------------------
+
+def _one_filing(concepts: dict, *, start: str, end: str, fp: str):
+    """A single 10-Q whose facts are the given {concept: value} at one YTD span."""
+    facts = {
+        name: {"units": {"USD": [_dur(start, end, val, "X", fp=fp)]}}
+        for name, val in concepts.items()
+    }
+    companyfacts = {"cik": 19617, "facts": {"us-gaap": facts}}
+    filings = {
+        "X": FilingMeta(
+            date.fromisoformat(end), date(2026, 5, 1), datetime(2026, 5, 1, 10, tzinfo=UTC), "10-Q"
+        )
+    }
+    return parse_snapshots(companyfacts, filings, {"X"})
+
+
+def test_revenue_reads_banks_total_revenue_tag():
+    # JPM/GS/WFC tag RevenuesNetOfInterestExpense in every 10-Q and never (or
+    # only annually) `Revenues` -- previously null, so revenue growth was too.
+    res = _one_filing(
+        {"RevenuesNetOfInterestExpense": 49836}, start="2026-01-01", end="2026-03-31", fp="Q1"
+    )
+    assert res.rows[0].revenue == 49836
+
+
+def test_revenue_reads_including_assessed_tax_variant():
+    # ARE/KHC tag only the Including variant.
+    res = _one_filing(
+        {"RevenueFromContractWithCustomerIncludingAssessedTax": 671},
+        start="2026-01-01", end="2026-03-31", fp="Q1",
+    )
+    assert res.rows[0].revenue == 671
+
+
+def test_revenue_concept_priority_is_unchanged_by_the_added_tags():
+    # The new entries are appended, so any issuer that already resolved keeps
+    # the same concept -- only issuers that resolved to nothing gain a value.
+    res = _one_filing(
+        {
+            "RevenueFromContractWithCustomerExcludingAssessedTax": 100,
+            "RevenueFromContractWithCustomerIncludingAssessedTax": 110,
+            "RevenuesNetOfInterestExpense": 120,
+            "Revenues": 130,
+        },
+        start="2026-01-01", end="2026-03-31", fp="Q1",
+    )
+    assert res.rows[0].revenue == 100
+
+
+def test_four_four_five_q3_ytd_span_is_accepted():
+    # A 12/12/12/16-week filer's YTD-Q3 is 36 weeks = 251 days (COST 2026 Q3),
+    # which missed the old 20-day tolerance around 273 by ~2 and dropped Q3
+    # every year -- breaking the quarter chain and nulling TTM and YoY.
+    res = _one_filing(
+        {"RevenueFromContractWithCustomerExcludingAssessedTax": 207431},
+        start="2025-09-01", end="2026-05-10", fp="Q3",
+    )
+    assert (date(2026, 5, 10) - date(2025, 9, 1)).days == 251
+    assert res.rows[0].revenue == 207431
+
+
+def test_eps_falls_back_to_continuing_operations_variant():
+    # REG tags only this variant on every filing; FCX tags it in its 10-K while
+    # using EarningsPerShareDiluted in its 10-Qs.
+    companyfacts = {
+        "cik": 910606,
+        "facts": {"us-gaap": {"IncomeLossFromContinuingOperationsPerDilutedShare": {
+            "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 1.81, "X", fp="Q1")]}
+        }}},
+    }
+    filings = {"X": FilingMeta(date(2026, 3, 31), date(2026, 5, 1),
+                               datetime(2026, 5, 1, 10, tzinfo=UTC), "10-Q")}
+    res = parse_snapshots(companyfacts, filings, {"X"})
+    assert res.rows[0].diluted_eps == 1.81
+
+
+def test_eps_concept_priority_is_unchanged_by_the_added_tag():
+    companyfacts = {
+        "cik": 831259,
+        "facts": {"us-gaap": {
+            "EarningsPerShareDiluted": {
+                "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 0.61, "X", fp="Q1")]}},
+            "IncomeLossFromContinuingOperationsPerDilutedShare": {
+                "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 0.75, "X", fp="Q1")]}},
+        }},
+    }
+    filings = {"X": FilingMeta(date(2026, 3, 31), date(2026, 5, 1),
+                               datetime(2026, 5, 1, 10, tzinfo=UTC), "10-Q")}
+    res = parse_snapshots(companyfacts, filings, {"X"})
+    assert res.rows[0].diluted_eps == 0.61
+
+
+# -- period identity from the fiscal calendar, not SEC's fy/fp ---------------
+
+from app.services.sec_facts_parser import _period_identity  # noqa: E402
+
+
+def _meta(end: str, form: str = "10-Q") -> FilingMeta:
+    d = date.fromisoformat(end)
+    return FilingMeta(d, d, datetime(d.year, d.month, d.day, tzinfo=UTC), form)
+
+
+def test_a_10q_is_never_labelled_fy():
+    # BXP: a 10-Q for period end 2026-03-31 carried fy/fp saying "2026 FY", which
+    # collided with the real annual row and measured a 90-day fact against the
+    # 365-day FY expectation.
+    fy, fp = _period_identity(_meta("2026-03-31"), "1231")
+    assert (fy, fp) == (2026, "Q1")
+
+
+def test_december_filer_years_do_not_collide():
+    # FRT: two 10-Ks, ending 2024-12-31 and 2025-12-31, both labelled "2024 FY".
+    assert _period_identity(_meta("2024-12-31", "10-K"), "1231") == (2024, "FY")
+    assert _period_identity(_meta("2025-12-31", "10-K"), "1231") == (2025, "FY")
+
+
+def test_january_year_end_groups_its_quarters(): 
+    # CRM/CRWD/WDAY: the year ending 2026-01-31 and its own quarters must share a
+    # fiscal year, and must not collide with the year ending 2025-01-31.
+    assert _period_identity(_meta("2026-01-31", "10-K"), "0131") == (2026, "FY")
+    assert _period_identity(_meta("2025-01-31", "10-K"), "0131") == (2025, "FY")
+    assert _period_identity(_meta("2025-04-30"), "0131") == (2026, "Q1")
+    assert _period_identity(_meta("2025-07-31"), "0131") == (2026, "Q2")
+    assert _period_identity(_meta("2025-10-31"), "0131") == (2026, "Q3")
+
+
+def test_mid_year_end_orders_correctly():
+    # STX: the year ending 2025-06-27 was labelled "2027 FY" and sorted after
+    # quarters that precede it.
+    assert _period_identity(_meta("2025-06-27", "10-K"), "0627") == (2025, "FY")
+    assert _period_identity(_meta("2025-10-03"), "0627") == (2026, "Q1")
+    assert _period_identity(_meta("2026-01-02"), "0627") == (2026, "Q2")
+    assert _period_identity(_meta("2026-04-03"), "0627") == (2026, "Q3")
+
+
+def test_four_four_five_quarters_place_correctly():
+    # COST: a 12/12/12/16-week year leaves Q3 112 days from the year end, not 91.
+    assert _period_identity(_meta("2025-11-23"), "0830") == (2026, "Q1")
+    assert _period_identity(_meta("2026-02-15"), "0830") == (2026, "Q2")
+    assert _period_identity(_meta("2026-05-10"), "0830") == (2026, "Q3")
+    assert _period_identity(_meta("2026-08-30", "10-K"), "0830") == (2026, "FY")
+
+
+def test_year_end_crossing_january_still_groups_one_year():
+    # DPZ (fiscalYearEnd 0102): the label shifts by one against Domino's own
+    # naming, which is fine -- a year and its quarters must simply agree.
+    year, _ = _period_identity(_meta("2025-12-28", "10-K"), "0102")
+    assert (year, "FY") == _period_identity(_meta("2025-12-28", "10-K"), "0102")
+    assert _period_identity(_meta("2025-03-23"), "0102") == (year, "Q1")
+    assert _period_identity(_meta("2025-06-15"), "0102") == (year, "Q2")
+    assert _period_identity(_meta("2025-09-07"), "0102") == (year, "Q3")
+
+
+def test_missing_fiscal_calendar_falls_back_to_filing_context():
+    assert _period_identity(_meta("2026-03-31"), None) == (None, None)
+    # ...and parse_snapshots then uses the fy/fp path, preserving old behaviour.
+    res = parse_snapshots(COMPANYFACTS, FILINGS, {"B"})
+    assert (res.rows[0].fiscal_year, res.rows[0].fiscal_period) == (2026, "Q2")
+
+
+
+def test_eps_falls_back_to_basic_only_when_no_diluted_variant_exists():
+    # PPL's 2026 Q1 tags no diluted EPS at all, only basic -- one missing period
+    # broke the quarter chain and nulled TTM.
+    companyfacts = {
+        "cik": 922224,
+        "facts": {"us-gaap": {"EarningsPerShareBasic": {
+            "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 0.60, "X", fp="Q1")]}
+        }}},
+    }
+    filings = {"X": FilingMeta(date(2026, 3, 31), date(2026, 5, 1),
+                               datetime(2026, 5, 1, 10, tzinfo=UTC), "10-Q")}
+    res = parse_snapshots(companyfacts, filings, {"X"}, fiscal_year_end="1231")
+    assert res.rows[0].diluted_eps == 0.60
+
+
+def test_diluted_still_wins_over_basic_when_both_present():
+    companyfacts = {
+        "cik": 320193,
+        "facts": {"us-gaap": {
+            "EarningsPerShareDiluted": {
+                "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 1.36, "X", fp="Q1")]}},
+            "EarningsPerShareBasic": {
+                "units": {"USD/shares": [_dur("2026-01-01", "2026-03-31", 1.40, "X", fp="Q1")]}},
+        }},
+    }
+    filings = {"X": FilingMeta(date(2026, 3, 31), date(2026, 5, 1),
+                               datetime(2026, 5, 1, 10, tzinfo=UTC), "10-Q")}
+    res = parse_snapshots(companyfacts, filings, {"X"}, fiscal_year_end="1231")
+    assert res.rows[0].diluted_eps == 1.36