Five Footnotes I Read Before I Trust a Balance Sheet

Picture two companies that both report $20 billion of long-term debt. One owes $1.5 billion a year for the next thirteen years. The other owes $6 billion in a single year, twenty-two months from now, and its bonds were sold when borrowing was cheap. The balance sheet prints the same line for both. Only one of them has a problem, and the only way to find out which is to open the notes.

I read footnotes before I read anything else about a company I do not know, and I have a fixed order for them. Debt maturities come first, then revenue recognition, leases and commitments, contingencies, and stock compensation. None of these notes is exciting. Together they tell me whether the totals above them can be trusted, and they cost about twenty minutes. Every company with a 10-K has them, and you can pull any filing from SEC EDGAR in a few clicks.

A word on numbers before I start. Every figure in the examples below is a round, made-up number chosen to show the arithmetic. None of it describes a real company, and I have avoided naming one, because the point is the method and because a footnote figure I have not checked this week is not worth repeating.

Debt due, year by year

The debt note contains a table that most readers skip: principal due in each of the next five years, and a lump for everything after. That table is the whole story. A company can carry a large balance and be perfectly safe if the maturities are spread out, and a modest balance can be dangerous if it is bunched.

Take the hypothetical $20 billion again. If $6 billion comes due in year two, management has to do one of three things: pay it from cash, refinance it, or sell something. Paying from cash only works if operating cash flow is large. Refinancing works if credit markets are open and the company’s rating has not slipped. If the old bonds paid 3 percent and new ones cost 6.5 percent, the interest bill on that $6 billion rises by $210 million a year until it is repaid. That is a permanent cut to profit made by a calendar, not by any decision about the business.

To size the question quickly, I divide long-term debt by the last four quarters of net income. The result is the number of years of current profit needed to repay everything. Four years is comfortable. Ten is heavy for an ordinary industrial company. Fifteen means the company is either a regulated utility, where customers repay the plants through rates and a high ratio is normal, or a business in trouble. I would skip the ratio for insurers and banks, which are built differently. I would distrust it when one large loss has depressed the four-quarter profit, because that inflates the result.

The ratio only tells me where to look. The maturity table tells me whether to worry. I also read the sentence near it about the revolving credit line, since an undrawn credit facility of, say, $4 billion is a cushion that changes how scary a $6 billion wall looks. And I check whether the facility itself expires before the wall does. That happens more often than you would expect.

When a sale counts as a sale

The revenue recognition note describes when the company books a sale. It is the dullest text in the filing and the one I am most reluctant to skip.

Several phrases matter. First, multi-year contracts: a company that signs a five-year, $500 million deal may count revenue as work is done, on billing, or up front for the license portion. Second, bundles: when hardware, software and support are sold together, the company decides how much of the price belongs to each piece. That split moves profit between years. Third, returns and rebates: a company that sells through distributors may book the sale on shipment and set aside a reserve for what comes back.

I read this note two ways. Once for what it says, and once against last year’s version. If the wording changed, the company has usually changed how it counts, and the note should explain why. A shift that pulls revenue forward by even 2 percent in a year when growth is 3 percent has done most of the growth by itself. Nothing in the income statement would warn you. I would want the explanation in plain language, and if the note is vague where last year’s was specific, that is a mark against management, not against the accountants.

The companion check is the gap between revenue and cash. If sales rise 10 percent and receivables rise 35 percent, customers are taking longer to pay or the company is recognizing sales it has not collected. Neither is a crisis alone. Both together, two years running, is a pattern I do not ignore.

Leases and promises not on the balance sheet

Leases used to hide. Accounting rules now put most of them on the balance sheet, but the note is still where you learn how long they run and what happens if the company wants out. A retailer with 1,000 stores on ten-year leases has committed to a fixed cost that does not shrink when sales do. Fixed costs work like this: when sales fall 8 percent, the rent does not.

I look at the table of future lease payments and compare it with annual operating income. If the next year’s lease bill is 40 percent of what the business earns before interest and tax, a modest sales decline will hurt a lot. I also look for purchase obligations, meaning fixed promises to buy materials or cloud capacity. A company that has committed $8 billion to suppliers while its own revenue is $12 billion has made a large bet on demand it cannot see yet. For businesses that need to build before they can bill, this is where the risk sits. I made the same argument about a big cloud backlog in an earlier note on Oracle, where the order book only becomes revenue after the buildout is paid for.

For a company with a fragile balance sheet, the lease and commitment notes are also the place to test a pattern I covered in my piece on GameStop’s balance sheet: cash looks strong on the front page, but commitments decide how much of it is really free.

Lawsuits, and how the company describes them

The contingencies note lists legal disputes and regulatory probes. Two words in it carry weight. “Probable” means the company expects to lose and has booked a charge. “Reasonably possible” means it has booked nothing and may give a range. The second category is where surprises come from.

I read the range. A company that says a reasonably possible loss is between zero and $300 million, on a business earning $2 billion a year, has told me the exposure is about 15 percent of a year’s profit. That is survivable. If the range runs to $3 billion, or the note declines to give one, I treat the stock as carrying an option that someone else holds against it.

I also compare this note with the previous year. Lawsuits that appear, then quietly disappear from the text are worth a search. Lawsuits that stay for five years with the same wording tell me management has no plan to settle. In healthcare and insurance, where claim costs drive everything, I read this note alongside the cost data, as I did in the piece on UnitedHealth’s medical cost ratio.

Footnote	What I look for	A warning sign
Debt	Principal due by year; credit line size and expiry	One year holds more than 25 percent of total debt
Revenue recognition	When a sale is booked; changes from last year	Wording changed, no reason given
Leases and commitments	Next-year payments against operating income	Fixed payments above 40 percent of operating income
Contingencies	“Probable” versus “reasonably possible”; the range	Range larger than a year of profit, or no range
Stock compensation	Shares issued against shares repurchased	Buybacks that only offset new grants

Buybacks that only pay the staff

The last note is stock compensation, and it answers a question the cash flow statement dodges: is the buyback shrinking the company or just holding it steady?

Suppose a company spends $3 billion a year repurchasing shares, and the share count at year end is the same as at the start. It has not returned $3 billion of value to holders in any way that raises earnings per share. It has paid its employees in stock and used cash to hide the dilution. Add the shares issued for pay, subtract the shares bought back, then look at the net. A net reduction of 3 percent a year is a real return. A net change near zero means the buyback is a cost of doing business dressed as generosity.

The note also gives the size of unvested awards and the expense recorded for the year. When stock pay is 12 percent of revenue, as it is at some software companies, the “adjusted” profit that leaves it out flatters the business by a wide margin. I always run one calculation: reported operating income minus stock compensation. If the company still looks good after that, I believe the story more.

How I use the five together

None of these notes is an instant sell signal. They are questions, and most companies answer them in an ordinary way. A steady maturity ladder and a plain revenue note, with a share count that actually falls: I read on and spend my time on the business. When one answer looks strained, I want a wider margin of safety in the price. When two or three do, I move on, because a cheap price is a poor reward for owning a company whose numbers depend on a refinancing or a court ruling.

I could be wrong about the thresholds. Twenty-five percent of debt in one year is a line I drew from habit, and a company with $10 billion in cash and a rock-solid credit line can carry a bigger wall than one with none. Context beats any rule in the table. What the notes give me is not a verdict but a list of things I would ask management if I ever met them.

If you want to start small, pick one company you already follow. Open its latest annual report and find the debt note. Write down the amount due in each of the next five years. If any single year is more than a quarter of the total, read the paragraph on how the company plans to meet it. That one number, checked once a year, is worth more than most of the commentary you will read about the stock.

Analysis and opinion only, not investment advice. Figures come from illustrative round numbers, not any company’s results; real notes sit in annual filings on SEC EDGAR and the SEC’s company search, and this method was last reviewed on September 22, 2026.