Vendor "revenue" for banks is a net construct, not the reported top line

Claim

For banks and lenders, vendor feeds do not report the income statement's top line. They report a net construct — approximately net interest income plus net fee income, i.e. revenue after funding cost. The issuer reports gross revenue. These are different numbers and the vendor does not label the difference.

Evidence

NU (Nu Holdings) FY2025:

Source	"Revenue"
Yahoo / roic.ai / `fin.py`	$10.63B
NU IFRS Total revenue (20-F)	$15,774.7M
NU Managerial Total revenue	$16,319.6M

The vendor figure understates the reported top line by 32.6%. Anything computed from it — P/S, EV/Revenue, revenue CAGR, net margin, revenue per share — is wrong, and wrong in a direction that makes the company look more expensive on sales and more profitable on margin simultaneously.

Second trap, same name: the geographic revenue table is a third basis

NU's IFRS 8.33 entity-wide disclosure totals $12.08B, not the $15.77B income-statement figure — it excludes gains/losses on financial instruments. Computing country mix against the headline gives ~70% Brazil; the correct answer is 91.3%.

Three different "revenue" numbers exist in the same filing set. State which basis you are using, every time.

Rule

For any bank, BDC, insurer or lender, take revenue from the primary filing and name the basis. If a report cites a revenue figure for a financial without saying whether it is gross, net, or the segment-note construct, the figure is not usable.

Related consequence: the same net/gross confusion makes cross-company comparison invalid unless both sides are on the same basis. A "P/S" comparison between a bank and a non-financial is meaningless regardless.

What would falsify this

A vendor labelling the basis explicitly, or reporting the issuer's gross top line. Re-check on the next financial analysed rather than assuming this persists.

[[principle-primary-source-beats-vendor]] · [[pitfall-roicai-fcf-field-returns-ocf]] — the adjacent failure, and note the pair: for a bank, both the revenue line and the cash-flow line from vendors are unusable. FCF is structurally meaningless for a lender anyway (operating cash flow is dominated by loan-book and deposit flows), so a bank analysis should reach for the primary filing from the start.

[[pitfall-yahoo-share-count-dual-class-fpi]] — the other trap found in the same analysis.

History

2026-07-31 — found in the NU analysis, alongside a 32% understatement of the loan-loss allowance in the same Yahoo payload ($4,165.7M reported vs $6,100M actual), which broke the coverage-ratio math until sourced from the 6-K.