The most expensive sentence Enron ever published did not appear in a press release. It sat in the related-party footnote of its annual report, where the company disclosed that it did business with partnerships whose managing member was "a senior officer" of Enron. Wall Street was reading the income statement, which showed revenue climbing past $100 billion in 2000. Jim Chanos of Kynikos Associates was reading the back of the filing. He wanted to know who sat on the other side of Enron's trades, why so much of its "earnings" came from marking long-dated contracts to model, and why its return on capital, by his own estimate, looked too thin to justify the stock's valuation. The shares peaked near $90 in August 2000. Enron filed for bankruptcy on December 2, 2001.
Chanos did not have inside information. He had the same 10-K everyone else had, and he read it in a different order. Short sellers often start where management least wants attention: the footnotes, then the accounting policies, then the numbers the company leads with. The fragility is rarely on page one. It sits in what quietly changed in the disclosures since the prior year.
Why the back of the filing tells the truth
Why does this matter? Because a 10-K is written by two authors with opposite incentives, and only one of them controls the front.
The 10-K is the annual report every US public company files with the SEC. Its opening sections, the business description and Management's Discussion and Analysis (MD&A), are where management frames the year. Executives choose which metrics to emphasize, which growth rates to quote, and which adjusted figures to lead with. That section is marketing that happens to be filed with a regulator.
The footnotes have a different author in practice: securities lawyers and auditors. Their job is not persuasion. It is protecting the company from liability. If a lawsuit follows a collapse, the defense rests on a simple argument: the risk was disclosed. That incentive forces material facts into the filing even when management would prefer they stay buried. The compromise companies reach is to disclose the fact while controlling its prominence. The related-party deal gets a paragraph, deep in the notes, written in passive voice.
So the economic mechanism is a split between emphasis and existence. Management controls emphasis. The law controls existence. A short seller's edge comes from ignoring emphasis entirely and auditing existence: what has the company been forced to say, and where has it chosen to say it quietly?
This split also explains why the market misprices the information for a while. Sell-side models are built from the income statement and management guidance. Few analysts rebuild their view from the notes each year, so a disclosure that is technically public can remain economically unpriced for quarters.
The practical point: treat MD&A as the company's argument and the footnotes as its testimony under oath. When the two disagree, believe the testimony.
The signal is the change, not the level
A single 10-K read in isolation is close to useless for this purpose. Every company has related-party transactions, judgmental estimates and contingent liabilities. The information is in the delta: what was added, removed or reworded relative to last year's filing.
Short sellers literally redline consecutive filings, running the prior year's text against the current one in a document comparison tool. Several sections reward this treatment.
Revenue recognition policy. This note describes when the company books a sale. A shift from recognizing revenue at delivery to recognizing it at shipment, or the addition of new language about "bill-and-hold" arrangements, means the same economic activity now produces earnings sooner.
Critical accounting estimates. This section lists the judgments that most affect reported profit: useful lives of equipment, allowances for bad debts, inventory reserves, warranty costs. Extend the assumed life of a server fleet from four years to six, and annual depreciation on that fleet falls by roughly a third overnight. Several large cloud providers made changes of this kind in recent years; none of it was fraud, but all of it lifted reported margins without a dollar of extra business.
Risk factors. Boilerplate risk language barely moves from year to year. When a new, specific paragraph appears (a named customer concentration, a covenant in a credit agreement, a regulatory inquiry), the lawyers saw something worth defending against.
Segment reporting. A company that reorganizes how it reports its divisions breaks the comparison with prior years. Sometimes the reorganization is genuine. Sometimes it buries a deteriorating unit inside a growing one.
Auditor changes and subsequent events. An auditor resignation, a late filing or a material weakness in internal controls is the filing's loudest alarm, and it is often the shortest paragraph.
The practical point: never read this year's 10-K alone. Put last year's next to it and read only what moved.
Where earnings and cash part ways
Accounting changes matter because they let reported profit drift away from cash. Profit is an opinion built from estimates; cash is a fact the bank confirms. The short seller's core quantitative test is whether the two still travel together.
The cleanest screen is days sales outstanding (DSO), the average number of days a company takes to collect what customers owe it.
DSO = (Accounts Receivable / Revenue) × 365
Take a hypothetical company. In year one it reports revenue of $1,000 million and receivables of $150 million.
DSO = ($150M / $1,000M) × 365 = 54.8 days
In year two, management celebrates 20% revenue growth to $1,200 million. Receivables have climbed to $260 million.
DSO = ($260M / $1,200M) × 365 = 79.1 days
Revenue grew 20%; receivables grew 73%. Customers are taking 24 more days to pay, or the company is booking sales it has not collected, or it has offered extended payment terms to pull demand forward. Every explanation points to the same conclusion: the growth in the headline is lower quality than it looks.
The companion test compares operating cash flow (the cash the business actually generated from operations) with net income. Over time a healthy company converts most of its reported profit into cash. When net income rises while operating cash flow stalls or turns negative, receivables and inventory are absorbing the difference.
Super Micro Computer is a recent illustration of the pattern. In its fiscal 2024 the company reported strong profit growth on surging AI server demand while its operating cash flow went negative as inventory and receivables ballooned. A delayed annual filing and an auditor resignation followed later in 2024. The cash gap alone did not prove wrongdoing; plenty of fast-growing hardware companies build inventory ahead of demand. It did tell a careful reader that the reported earnings rested heavily on balance-sheet items that had not yet turned into money.
The practical point: compute DSO and the cash-to-earnings ratio for every name you own, every year. The calculation takes two minutes from the balance sheet and cash flow statement, and a widening gap tells you where in the footnotes to look next.
When the backwards read misleads
The method has real failure modes, and they mostly come from mistaking an innocent change for a hostile one.
New accounting standards force disclosure changes across entire markets at once. When ASC 606 rewrote revenue recognition rules in 2018 and ASC 842 moved operating leases onto balance sheets in 2019, nearly every filer's notes changed. A redline that year flagged everything and therefore signaled nothing. Check whether a change was company-specific before treating it as information.
Expanded disclosure can also be a sign of health. Companies frequently add detail after receiving an SEC comment letter, the regulator's request for clarification. A longer risk factor section sometimes reflects a more careful legal team, not a more fragile business.
Growth itself distorts the cash tests. A company landing large enterprise contracts can see DSO rise because big customers negotiate longer payment terms. The question is whether the trend persists and whether the allowance for doubtful accounts, the reserve for receivables the company does not expect to collect, is shrinking as a percentage at the same time. A rising DSO with a falling reserve is the combination that should worry you.
Two limits are more fundamental. Outright fabrication can survive a clean read of the notes. Wirecard's annual reports carried auditor sign-off for years while roughly €1.9 billion of reported cash did not exist; that collapse was exposed by investigative reporting and third-party confirmation, not by footnote language. And being right is not the same as being on time. Chanos's Enron thesis took roughly a year to pay, and many correct short theses lose money because the market ignores the footnote longer than the short seller can fund the position.
The practical point: a changed disclosure is a reason to investigate, not a conclusion. Rule out the industry-wide and benign explanations before you act on it.
Reading in the short seller's order
You do not need to bet against a company to use this method; owners benefit from it more than anyone. The discipline is a reordering of attention.
Start with the notes on related parties, contingencies and subsequent events, because those are the disclosures management most wants minimized. Move to the accounting policies and critical estimates, comparing them line by line with the prior year. Then run the cash tests on the statements. Only after that read MD&A, and read it as an argument to be checked against everything you have already found. If management leads with an adjusted metric, find the reconciliation to the GAAP figure and see which adjustments grew fastest.
The mechanism reduces to one sentence: management decides what you notice, but disclosure law decides what exists in the filing, so the gap between the two, measured year over year, is where mispricing lives. A headline number tells you what the company wants priced. A changed footnote tells you what the market has not priced yet, and that second piece of information is the one with value.
