Yahoo's snapshot "freeCashflow" field disagrees with a reconciled sum of its own quarterly statements

Claim

/analyze-from-before DPZ reconciled TTM free cash flow two independent ways from the quarterly cash-flow statements (Q3 2025 through Q2 2026):

Quarter	OCF	Capex	FCF (quarter's own field)
Q3 2025	$185.40M	$21.44M	$163.96M
Q4 2025	$239.81M	$63.89M	$175.92M
Q1 2026	$161.96M	$15.04M	$146.91M
Q2 2026	$190.65M	$23.97M	$166.68M
TTM sum	$777.81M	$124.34M	$653.47M

OCF − Capex = $653.47M. The sum of each quarter's own pre-computed "Free Cash Flow" cell = $653.47M. Both methods agree to the dollar, using only data returned by the same MCP server. get_stock_info's own totalRevenue and operatingCashflow snapshot fields for DPZ do match this quarter-sum (operatingCashflow: 777,804,992 ties to $777.81M above). But the snapshot's freeCashflow field returned $531,496,512 — 19% below the reconciled $653.47M.

Why this is not the known single-quarter-read-as-TTM pitfall

[[pitfall-single-quarter-fcf-read-as-ttm]] documents an analyst misreading one quarterly column as a trailing year, on a seasonal business (Intuit) where one quarter plausibly resembles a full year. This is different: DPZ is not strongly seasonal, no single quarter's FCF ($147-176M) is anywhere near the discrepancy's size, and the vendor's own aggregated field — not an analyst's reading of it — disagrees with the vendor's own quarterly data. The mechanism is unconfirmed (possibly a different capex definition, a stale computation window, or a different set of trailing quarters than the ones the quarterly endpoint returns), but the effect is the same direction as every other vendor-field trap in this playbook: the vendor number is smaller than the reconciled truth, which would understate FCF/share and inflate an EV/FCF or P/FCF multiple if quoted directly.

The rule

1. Never quote get_stock_info's freeCashflow (or any single-field trailing aggregate) for DPZ, or any name, without reconciling it against a quarter-sum first.
2. The reconciliation is cheap — get_financial_statement(quarterly_cashflow) already returns a pre-computed "Free Cash Flow" cell per quarter; summing four of them costs nothing extra beyond the call already needed for other quarterly data.
3. If the snapshot field and the quarter-sum disagree by more than a rounding difference, use the quarter-sum — it is derived from the same underlying statement data with a visible, checkable computation, where the snapshot aggregate is not.
4. Confidence upgraded to high on 2026-08-27 — 13 confirmed instances in one batch (see the amendment below), from micro-cap (SMR) to mega-cap (AAPL, $137B). This is the field's normal behaviour, not an edge case. The snapshot freeCashflow field is not usable; recompute from the quarterly statements every time FCF is verdict-bearing.

Related

[[pitfall-single-quarter-fcf-read-as-ttm]] — the neighbouring failure at the analyst layer rather than the vendor layer. [[pitfall-roicai-fcf-field-returns-ocf]] — a different vendor, a different FCF-field defect, same governing rule: a vendor's "free cash flow" field is not self-verifying. [[principle-primary-source-beats-vendor]].

History

• 2026-08-27 — found during /analyze-from-before DPZ, while reconciling TTM FCF ahead of a P/FCF valuation call.
• 2026-08-27 (batch) — the same-day /analyze + /analyze-from-before batch confirmed the field fails on 13 names across the whole cap spectrum, upgrading confidence to high and mapping two distinct failure modes:

Ticker	snapshot freeCashflow	reconciled quarter-sum	miss	note
DPZ	$531.5M	$653.5M	−19%	baseline case
GPN	$7.87B	$822.7M	9.6×	acquirer (Worldpay/Issuer swap)
LEU	−$127.3M	−$163.8M	−$36M	capex ramp
GWRE	$283.2M	$319.6M	−11%
CRWD	$1.926B	$1.431B	+35%	over-states (rare direction)
PANW	≈−6%		−6%	acquirer (CyberArk)
PCTY	$359.9M	$427.8M	−16%
ISRG	$2.615B	$3.223B	−19%
AAPL	$107.7B	$136.7B	−21%	mega-cap
BWXT	$145.6M	$316.8M	−54%	acquirer
SANM	−$222.2M	+$594.1M	sign flip	acquirer (ZT Systems)
CEG	≈−$6.6B	+$0.31B	sign flip	acquirer (Calpine)
SMR	−$204.6M	−$778.8M	frozen	mechanism 2

Mechanism 1 — M&A-quarter mis-netting. The largest misses (BWXT −54%, and the SANM/CEG sign flips) all sit on names with a big acquisition inside the trailing four quarters — the purchase-of-business investing outflow appears mis-netted into the aggregate. Standing rule: whenever a large M&A quarter sits in the trailing four, recompute FCF from the quarterly statements — never trust the snapshot field for an acquirer.

Mechanism 2 — the field freezes at a stale single quarter. On SMR the snapshot freeCashflow (−$204.6M) reproduced Q4 2025's single-quarter FCF (−$204.07M) almost to the dollar, while the sibling operatingCashflow field on the same snapshot object was correctly TTM-summed. The two aggregate-looking fields on one object can sit on different windows — one TTM, one frozen at a prior quarter. This overlaps [[pitfall-single-quarter-fcf-read-as-ttm]] but at the vendor layer: the vendor, not the analyst, served one quarter as if it were the year.

Direction is not even consistent (CRWD over-states +35% while everything else under-states), so the field cannot be corrected with a fudge factor — only replaced by the quarter-sum.