Skip the packaging, read the filing
Read the real filing on SEC EDGAR, not the glossy shareholder letter. The 10-K is a legal document: management can spin, but outright lying is a crime, which makes its language unusually informative.
Section 1 — Item 1: Business (how they actually make money)
What to extract: what the company sells, to whom, how it charges (one-off vs recurring), segment mix, and customer concentration. Red flag: "one customer represented 22% of revenue". Green flag: revenue spread across many customers, recurring contracts, switching costs described in plain terms.
Section 2 — Item 1A: Risk Factors (read the deltas)
Most risk factors are boilerplate ("markets may decline"). The gold is in what changed vs last year: new risks appear because lawyers made them — something real happened. Compare year over year; a newly added paragraph about "supplier dependence" or "regulatory inquiries" is the earliest public warning you'll get.
Section 3 — Item 7: MD&A (management explains the numbers)
Management's Discussion & Analysis is where the story meets the accounts: why revenue moved, what's driving margins, liquidity and capital plans. Cross-check it against the financial statements — when the narrative says "strong demand" but receivables grew twice as fast as revenue, the numbers are telling a different story than the prose. That gap is exactly where mispricing lives.
A 15-minute routine
- Minutes 1–5: Item 1 — write one sentence: "This company makes money by ___".
- Minutes 5–8: Item 1A — only the risks that are new or expanded this year.
- Minutes 8–13: Item 7 — revenue/margin drivers; check them against the income statement and cash flow.
- Minutes 13–15: footnotes scan — debt maturities, leases, dilution, litigation.
Where AI genuinely helps (and where it doesn't)
AI is excellent at the first pass: locating the sections, summarising 60 pages of prose, and flagging year-over-year changes in risk language. It should not be trusted blindly on numbers — good tools cross-check the narrative against as-filed fundamentals rather than paraphrasing the PDF. That's how we built it into Compounder: the AI reads Business, Risk Factors and MD&A straight from EDGAR and contrasts them with the reported figures.