Home / Guides
Written from the bugs, not from a textbook
Guides to reading a filing without getting a wrong number
Every one of these started as something that broke in our own code, got traced back to a filing, and turned out to be a trap anyone building a valuation from SEC data walks into. The figures in them are read live from EDGAR and the scripts that produced them are published alongside, so nothing here has to be taken on trust.
Which XBRL tag is free cash flow, and why the obvious one is wrong
Free cash flow is two tags subtracted, and both have aliases that large companies quietly abandoned years ago. The abandoned tag keeps answering. Amazon's returns $6,737m from 2016 next to the real $131,819m from 2025, and AT&T's stale figure is within one percent of the right one, which is far more dangerous. A survey of twenty large filers, with the rule that fixes it.
14 October 2026 · 20 companies · evidence
Shares outstanding or diluted weighted average: which one goes in a DCF
Two tags, two different questions. The cover-page count is a point in time; the diluted weighted average is a trailing-year figure that already includes dilution. Two of twenty large filers — Alphabet and Meta — never tag the cover-page count at all, which is exactly the gap that once carried Nvidia's 24.3bn-share count into a Shopify valuation and priced it at €5.10 instead of €94.97. A survey of twenty large filers, with the rule that fixes it.
19 October 2026 · 20 companies · evidence
What to do when terminal growth is above the discount rate
The terminal value formula is only defined for r > g, and nothing in the arithmetic enforces that. Carry a company's own trailing growth straight into the terminal year and Chevron's naive 5-year free cash flow growth comes out at 58.57%, JPMorgan's ordinary-looking 8.75% revenue growth still prices a 435× terminal multiplier. Half of twenty large filers read live from SEC EDGAR already break the formula or come within one bad assumption of it.
26 October 2026 · 20 companies · evidence
Net debt: why adding up every debt tag overstates it
US-GAAP debt tags nest: LongTermDebt already holds its current part, and
DebtCurrent already holds commercial paper and short-term borrowings. Add up every tag a
balance sheet answers and AT&T's debt comes out at $279.8bn instead of $134.7bn, while Amazon and
Nvidia turn from net cash into net debt. For 13 of 18 comparable large filers read live from SEC EDGAR,
the naive sum is at least 1.8 times the real figure.
2 November 2026 · 20 companies · evidence
Next
One guide a week. Coming up: cash, and why CashAndCashEquivalentsAtCarryingValue on its
own can understate what a company has to net against its debt. Pfizer's is $1.1bn. Read live and checked
across the same large filers.
Rather use the arithmetic than read about it? The calculator applies these rules to any SEC filer and names the tag behind every figure it fills in.
Open the calculator