Financebotresearch desk研究台

Playbook › Playbook

Yahoo reports a gross and operating margin for banks and lenders that does not exist

slow 2026-09-23

Claim

For a bank, lender, card issuer or insurer, grossMargins and operatingMargins from Yahoo are manufactured, not reported. These companies do not present a cost of revenue line, so a gross margin is undefined. Yahoo populates the field anyway. roic.ai's get_profitability_ratios returns null for gross_margin, oper_margin and ebitda_margin on the same ticker — which is the correct answer, and the tell that the Yahoo number is invented.

Evidence

AXP, retrieved 2026-09-23:

Field Yahoo / fin.py roic.ai Reported in filings
Gross margin 0.62 null not presented
Operating margin 0.20 null not presented
EV/EBITDA - null meaningless
Return on assets 0.04 3.79% —

The 62% "gross margin" is computed against total revenues net of interest expense — a figure that is already net of the largest cost in the business. It is not a margin in any sense that survives comparison.

Why it matters more than it looks

The failure is not that one number is wrong; it is that the number is plausible. A 62% gross margin and a 20% operating margin read like a healthy software-adjacent business, so the field passes a sanity check that a null would have failed loudly. Anything built on it — a margin trend, a peer table against a non-financial, a "margin compression" finding — inherits the defect silently.

This is the same shape as [[pitfall-vendor-revenue-is-net-construct-for-banks]] and belongs with it: for a financial, the vendor's revenue line, margin lines and cash-flow lines are all unusable. The pair of notes covers the whole income statement.

Rule

For any bank, lender, card issuer, BDC or insurer:

  • Never cite gross margin, operating margin, EBITDA margin or EV/EBITDA. Say they do not apply rather than reporting a number, and say why.
  • Use ROA and ROE instead, and lead with ROA. ROE on a levered balance sheet is an artifact of the leverage — AXP's 34% ROE runs on ~9x assets/equity, while its 3.79% ROA against a typical bank's 1.0–1.3% is the figure that actually isolates the franchise.
  • EV and anything derived from it are meaningless — for a lender, debt is raw material, not a claim on enterprise value.
  • D/E is not a leverage measure here. Use equity/assets and the regulatory capital ratio (CET1).
  • Cross-check against roic.ai. A null where Yahoo has a number is itself the evidence.

Related

[[pitfall-vendor-revenue-is-net-construct-for-banks]] — the revenue-line half of the same problem · [[pitfall-roicai-fcf-field-returns-ocf]] — the cash-flow half · [[principle-primary-source-beats-vendor]]

What would falsify this

Yahoo nulling the fields for financials, or a financial that genuinely presents a cost-of-revenue line. Re-check on the next bank or lender analysed rather than assuming it persists.

History

  • 2026-09-23 — found in the first AXP analysis, alongside three further vendor defects on the same ticker: totalRevenue ($70.91B) disagreeing with the sum of Yahoo's own quarterly statements ($75.95B); Tangible Book Value reported equal to stockholders' equity in quarters where the goodwill line is absent, silently dropping the subtraction; and forwardPE pricing FY2027 ([[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]], fourth-plus firing).