How the numbers are calculated

Last updated: October 16, 2026

Leftagraph charts what companies file with the SEC. Most numbers are shown exactly as filed. Some we work out. This page says which, how, and how we check them.

Where the numbers come from

• Financial statements: the SEC's public EDGAR database. Companies tag each figure in their 10-K and 10-Q reports (the format is called XBRL), and the SEC publishes those tagged figures as "company facts".
• Statements as filed: the SEC's Financial Statement and Notes data, which keeps each balance sheet, income statement and cash flow statement the way the company laid it out: its lines in order, in its own words, including lines on the company's own tags, and which lines add up to which total. Total debt and liquid cash are read from the balance sheet; revenue, net income, operating cash flow, capital spending, depreciation & amortization and dividends from the income and cash flow statements. See Total debt and liquid cash and Revenue and the cash flow lines.
• Financial reports only. Numbers are read from a company's annual and quarterly reports and their amendments, recast statements it files with the SEC, and, for the years before its first annual report, its registration statement. Not from proxy statements: their executive-pay table repeats five years of net income, and a mistake in one could overrule the annual report.
• US accounting rules only. Companies that report under international rules (most foreign filers) are not covered yet.
• Share prices (Pro): the IEX exchange. See Share prices.

How we read a filing

• Read from the statements themselves. Revenue, net income, operating cash flow, capital spending, depreciation & amortization, dividends, total debt and liquid cash are read from the company's own statements, line by line (below).
• The other lines: one line, several tags. Companies use different tags for the same figure, and switch over time. Each of the other lines reads a short list of tags in a fixed order, and the first one with a value for a date is used.
• Restated numbers. When a company reports the same period again in a later report, the most recently filed version is used.
• What counts as a year or a quarter. A year is a period of 350 to 380 days. A quarter is 80 to 100 days. Other period lengths are not charted on their own.
• A fiscal year, not just any twelve months. Some companies print a "twelve months ended" column in every quarterly report. Those are not fiscal years. A twelve-month figure is charted as a year only when an annual report carried it (or, before a company's first one, its registration statement).
• Balance sheet dates. Balances (cash, debt, equity) are shown only for dates that end one of the company's fiscal years or quarters.
• Period names. A period is named by the month it ended ("Jun 2025"), not by the company's own quarter number.

Quarterly numbers

Companies often don't file a figure for a single quarter. A 10-Q gives cash flow for the year so far, and a 10-K gives the full year but not the fourth quarter. Where the three-month figure is filed, it is shown as filed. Where it isn't, we work it out by subtraction:

• Year-so-far figures: a quarter is one cumulative figure minus the one before it, when both start on the same date. Nine months minus six months is the third quarter.
• The fourth quarter: the full year minus the three quarters inside it.

This is exact arithmetic on filed numbers. A quarter worked out this way is marked calculated.

Not done for earnings per share, dividends per share or share counts. Per-share figures and averages can't be subtracted safely: a special dividend can be in the year's figure and not in the quarters', and subtracting would put it in the wrong quarter. So the fourth quarter of those lines usually shows "no data".

Total debt and liquid cash: read from the balance sheet

Debt and cash are filed under dozens of different tags, and a list of tags can only guess which of them are on a company's balance sheet. So these two lines are read from the balance sheet itself, the way the company laid it out:

• Total debt is every borrowing line on the balance sheet: debt, borrowings, notes and loans payable, commercial paper, credit facilities, mortgages. That includes a line on the company's own tag, which the SEC's "company facts" leave out (a property trust's "revolving credit facilities and commercial paper", say). Leases are not debt here, as before; where a company shows debt and finance leases as one line, that line is counted as the company shows it.
• Liquid cash is the cash and investment lines in current assets. A balance sheet with no current assets (a bank, an insurer, a property trust) has no "current" to go by, so there it is cash and the lines that say short-term, such as short-term investments or Treasury bills. Restricted cash is not liquid and is left out.
• Nothing is counted twice. The filing says which lines add up to which total. A line and its own total are never both counted, and a total that holds something we leave out (cash plus restricted cash) is counted through its parts. Where the filing doesn't link a subtotal, one equal to the lines right above it is counted once.
• Balance sheets shown per business. A few companies show some lines only per business, side by side (a car maker and its finance arm, a conglomerate's insurance and railroad sides). Those lines are the sum of the businesses the balance sheet shows, used only where they add up to the company's own totals.
• Only a balance sheet that adds up is read. Its totals (assets, current assets, liabilities, current liabilities) must equal their parts. A filing that doesn't is passed over for another that shows the same date.
• Which filing. A year-end appears in that year's annual report, in the quarterly reports after it in short form, and in next year's annual report. The annual report is used, the most recent one first, so a restated figure wins as it does everywhere on the chart.
• The newest filing. The SEC publishes this data once a month, a week to five weeks after the filings in it. Until it has a company's newest report, that report's balance sheet is read with the same lines as the company's last one and the numbers it filed under those lines' tags, where that is sure to give the same reading: not where a line counted is on the company's own tag or shown only per business, and only where every total on it still adds up, so a line the newer report added can't be missed.
• No balance sheet read, no number. Where there is no balance sheet to read for a date (before 2009, a filing that doesn't add up, a company the SEC's data doesn't have yet), total debt and liquid cash show "no data", and so does everything built on them. Nothing from a list of tags or a formula stands in for the company's own balance sheet.
• The checks still run on every figure: total debt against the interest bill, liquid cash against current assets, and the bank rule (all below).

Revenue, net income and the cash flow lines: read from the statements

These lines used to be read from a list of tags, and a company that filed under a tag the list didn't have got "no data", or a figure from its notes. So, like total debt and liquid cash, they are read from the company's own income and cash flow statements, line by line. One rule for every company:

• Revenue is the revenue total on the income statement. Where it shows more than one revenue line, the total wins over its parts, in the SEC's own order: total revenues, then revenue from customer contracts, then net sales. A bank or broker with no total revenues line has its revenue net of interest expense. Product sales and service sales are added up only where there is no total. A bank that prints no revenue total at all has its net interest income and its total non-interest income added up, marked calculated: that sum is what banks that do print one call total revenue. On a bank's statement a fee line such as service charges on deposits is part of its non-interest income, never its revenue.
• Net income is the net income for the company's own shareholders. Where the statement shows none, it is the consolidated net income less the minority holders' share, both from the same statement. A company with nothing discontinued that tags its net income line as income from continuing operations has that line read, since it then is the net income; so does a company whose own line is worded plainly "Net income".
• Operating cash flow is the operating total on the cash flow statement.
• Capital spending is the money spent buying property, plant and equipment, or software, in the investing section. Not property bought as an investment (a property trust's purchases of buildings; its improvements to them do count), not a bank's foreclosed property, not businesses, not securities. Where a company shows several such lines, they are added up. Money from selling equipment is not subtracted, and a line shown "net of" grants or the like is still spending.
• Depreciation & amortization is the operating section's lines for depreciation, depletion and the amortization of intangible assets and software, added up where the company shows them separately. Not the amortization of loan costs, bond discounts or other deferred items, and not a tenant's leases. A company that owns equipment and rents it out (a carmaker's finance arm leasing cars) depreciates what it owns, and that does count.
• Dividends are the financing section's dividends and distributions to the company's own shareholders, common and preferred, or its partners. Not payouts to minority owners of its subsidiaries.
• Money in or money out. A cash flow line's direction is read from the statement's own arithmetic: each section adds up to its total. We use the signs the company shows where they make the section add up, otherwise the filing's links between lines, otherwise what the line is (a payment is money out). So capital spending and dividends are always the amount paid, even where a company's tags carry the opposite sign.
• Sections. A company that lists its cash receipts and payments (the direct method) also reconciles its net income to the same operating total below the statement. Both parts are its operating section, so its depreciation is found there.
• Which filing. The company's own annual and quarterly reports, the most recently filed first, so a restated figure wins. A filing whose statement in the SEC's data turns out to be another table (it happens) is skipped for that statement: its periods come from another filing, or are read as the newest report is. A registration statement or another filing only for a period none of them shows: such a filing can carry the statements of a business the company bought, and those are not its own.
• The newest report. Until the SEC's monthly data has it, a report is read with the lines of the company's last statement and the numbers it filed under those lines' tags. Where a line is on the company's own tag, that period shows "no data" until the monthly data has it. A year the monthly data lacks, between years it has, is read the same way.
• No statement, no number. Where there is no statement to read for a period, or the statement has no such line, the line shows "no data", and so does everything built on it. Nothing from a list of tags stands in.
• Marked when added up. A figure read from one line is the company's own; one added up from two or more lines (capital spending on property plus software, say) is marked calculated.

Numbers we calculate

There are two kinds.

Fill-ins: the company's own figure when filed, calculated when not

Five lines: Gross profit, Short-term debt, Long-term debt, Operating lease liabilities, and Total liabilities. (Net income is worked out the same way where a company shows none, but from the lines on its own income statement: see below.)

• The company's own figure always wins. The formula is used only on dates where the company filed no figure for that line.
• The formula uses only numbers the company filed for that same date. Each formula is in the list below.
• A part shown in italics in the list counts as nothing only for a company that files that part on no date at all: a company with no goodwill files no goodwill line. A part the company files in other periods but not in this one is unknown for this one, so the sum shows "no data" there rather than a figure that leaves it out.
• How often the formula is used differs by line. Few companies file short-term debt as a single figure, so on most charts it is calculated. Most companies file the other five themselves.

Depreciation & amortization: from the cash flow statement

Read from the cash flow statement (above): one line as the company shows it, or the company's depreciation and amortization lines added up where it shows them separately, marked calculated. On the statement nothing is hidden: those lines are the whole figure. (Until October 10, 2026, it was shown only where a company filed one total, because adding up the tags we could find left out parts the company's own total held.)

Our own lines: always calculated

Lines no company files as one figure: net cash, total debt (the sum of the borrowing lines on the balance sheet), free cash flow, owner earnings, working capital, tangible equity, book value per share, and every margin, return, ratio, yield and valuation multiple. A ratio is left blank when its bottom number is zero or negative, rather than showing a number that means nothing.

One check: total debt against the interest bill

Total debt is read from the balance sheet (above), which finds debt under any tag. The check stays as a safety net, for a borrowing line worded in a way we don't recognise: debt we missed would make total debt far too low and net cash come out as plain cash. A wrong net cash figure is worse than none, so the filing is made to check itself:

• The check. The interest a company paid in a fiscal year is compared with the debt we read at that year-end, plus its finance leases. Read correctly, interest is a few percent of debt. When it is more than 25%, part of the debt is somewhere we didn't read.
• What happens then. Total debt and net cash are left blank for that year and its quarters, and so is everything built on them: debt to equity, years to pay off debt, return on invested capital. The chart shows "no data". Cash, the debt lines the company did file, and every other line are untouched.
• When it is skipped. When the interest is under 2% of liquid cash, it is most likely fees or lease interest at a company with little or no debt, and the number is kept. A company that files no interest figure can't be checked this way.
• Parts can't exceed the whole. Total debt more than the company's total liabilities means something was counted twice or read wrongly. That year is left blank too, interest or not.
• What it costs. A company that pays off most of its debt during a year has a full year's interest against a small year-end balance. That one year is left blank although it was read correctly.
• The company's own figure for all its debt, from the notes to its accounts, is not used. Total debt is what the balance sheet shows.

Every sum is checked against the company's own total

When a company doesn't file one of these figures, we add it up from its parts. Each sum is then checked against a total the company does file, for the same date. A part can't be bigger than the whole:

• Liquid cash can't be more than current assets.
• Short-term debt can't be more than current liabilities.
• Long-term debt and operating lease liabilities can't be more than total liabilities.
• Gross profit can't be more than revenue.
• When a sum is bigger, that date is left blank ("no data" on the chart), and so is everything built on it that date, net cash and total debt included. A number we worked out that doesn't fit the company's own totals is a number we read wrongly, and a wrong number is worse than none.
• Figures the company filed itself are never changed by this. Where it has no total to check against (banks and insurers often file no current assets), the sum is shown unchecked.

Banks: no total debt or net cash

A bank's cash is mostly money its customers have deposited, and those deposits are not debt. Cash minus debt says nothing about a bank's strength, so we don't show it.

• What is left out. Total debt and net cash, and everything built on them: debt to equity, years to pay off debt, return on invested capital and EV/EBIT. They are left out for every year, and a line under the chart says so. The debt lines the bank filed, its cash and every other line are shown as usual.
• Who counts as a bank. A company whose customer deposits, at its latest fiscal year-end, are at least 25% of everything it owes and at least $100 million. Some small companies that aren't banks file a figure called "deposits"; the $100 million keeps them out. Where a bank files its deposits only in two parts (those that earn interest and those that don't), the two are added; both parts have to be filed.
• One answer per company. It is decided from the latest annual report and applies to every year, so a bank's chart never shows net cash for some years and not others.
• Card issuers, brokers and other companies that own a bank count when that bank's deposits are a large part of what they owe, because then their cash is mostly customers' money too. A company with a small bank inside it (a motorcycle maker's finance arm, say) doesn't, and keeps its numbers.
• In the screener, a bank's total debt, net cash and the lines built on them are empty, so a filter such as "net cash of zero or more" never lists a bank.

One rule with a tolerance: stock held in other companies

A company's stock holdings carry the same SEC tag whether they are short-term or long-term, and the SEC data doesn't keep the balance sheet's layout. So stock holdings count as liquid cash only on dates where the company's own current assets add up with them, to within 0.1%. That shows the company listed them as a current asset. When it can't be shown, they are left out, so liquid cash can read low but not high.

How a calculated number is marked

• On the chart: a dashed line. A line that is always calculated is dashed from end to end. A line that is normally filed is dashed only around the periods we filled in. A solid line is as filed.
• Also on the chart: a hollow point (○) when the chart is zoomed in far enough to draw points, and the word "calculated" beside the value when you point at it.
• A key under the charts shows a sample of each kind of line on screen.
• In the data table: an asterisk (*) and italics. A line that is always calculated carries the asterisk on its column heading.
• In a CSV download: the numbers stay plain numbers, and a column at the end names the calculated lines in each row.

Missing periods

• When a line has no number for a period, the chart draws a dotted stretch across it and shades the period with a "no data" flag. The data table shows a dash. Dotted means nothing was reported; dashed means we calculated it.
• A line with no number in any of the years shown says "no data" in its label above the chart.
• "No data" doesn't mean "not filed". The company usually did file that report. This one figure isn't in the SEC's standard data for it: it was left out, or filed under the company's own tag. We can't tell which, so we don't claim to.
• A figure a company reports only in some years (buybacks, for example) shows "no data" for the years between. Often the real amount was zero. The filing doesn't say so, so neither do we.
• Periods before a line's first number are not flagged. The figure may not have existed yet.

Stock splits

After a stock split, older per-share numbers no longer line up with today's shares. A company restates only the last few years itself, so a line drawn exactly as filed falls off a cliff at the split: earnings per share seems to drop by three quarters in a year when profit rose. So per-share figures and share counts are put on today's share basis on every chart, and so is the share price where it is shown (Pro). Each number changed this way is marked split-adjusted (a † in the data table).

• Where the split list comes from: the company's own filings. After a split, a company restates earlier periods, so the same period appears twice with share counts a clean ratio apart (2, 3, 4, 10…).
• Which numbers are changed: each number's filing date says which share count it was reported on. A number last filed before the split is changed. One the company restated itself is left exactly as filed.
• A split is accepted only when two or more different figures in the filings agree on the ratio and at least one of them is a share count. Where share prices are loaded (Pro), a split of 2-for-1 or larger must also show in the price: the price has to have moved by that ratio.
• When it is dated. With share prices, to the day the price moved. Without them, to the filing that first shows the new share count.
• Deliberately cautious. A missed split leaves a visible jump in the line. An invented one would bend real data.
• Not detected: stock dividends smaller than 6-for-5, and a split so recent that no filing reflects it yet. That one shows as a sudden jump until the company's next report.

Share prices and valuation (Pro)

• The price shown is the last trade on the IEX exchange in regular trading hours (IEX's own closing auction price where it runs one). It is a day behind. It is not the official closing price, though for widely traded shares the two are usually within cents.
• One point per month (the last trading day we have), plus the most recent price.
• A thinly traded share may have no IEX trade on a given day. Then the latest one we have is shown, with its date.
• Price next to yearly figures. P/E, P/B, market cap and the yields pair each month's price with the latest full fiscal year that ended on or before that month. P/E here is price on the last full year's earnings, not the last twelve months.
• Ticker changes. Prices are tied to the company, not just the ticker symbol, so a ticker another company used to have doesn't bring that company's old prices with it. For a company that changed its ticker, the price line can start at the change.

Data provided for free by IEX. By accessing or using IEX Historical Data, you agree to the IEX Historical Data Terms of Use.

How we check our work

• Fill-ins and subtracted quarters are arithmetic, not estimates. Every input is a number the company filed, for the same date or the same fiscal year. You can redo any of them by hand from the filing.
• Automated tests run on every change to the site, and a change that fails one can't go live. They cover tag reading, restatements, quarter subtraction, each kind of formula, split detection and the missing-period flags.
• A nightly check for dropped inputs. Across every company, it flags any figure a formula depends on that was filed a year ago and isn't now. That is the sign of a company changing how it tags something.
• A nightly check of split detection against a list of well-known splits, and of companies known to have had none.
• A nightly price check. Our prices for a fixed set of 30 companies are compared with official closing prices from a separate source. A gap of more than 1% is flagged for a person to look at.
• A nightly count of figures from filings that aren't financial reports (proxy statements and the like) that differ from the company's annual or quarterly report. They are never read; the count shows how often one would have changed a number.
• A nightly count of what the checks left blank. The total-debt check, the bank rule and the parts-and-whole check each report how many companies they touched and the largest of them, so a check that is too strict, or a new kind of mistake, shows up the next morning.
• You can tell us. Under every chart, "Number look wrong? Tell us" opens an email with the company, the chart and its lines already filled in.
• A nightly comparison with the company's own figure. Some companies file both a figure and its parts. For those, we run our formula anyway and compare it with what the company filed. A drop in how often a line matches, or one cause showing up at many companies, is flagged automatically.

What we can't confirm. Where a company files only the parts, there is no figure of its own to compare with, so a fill-in can't be confirmed for that company. And where companies do file both, our formula does not always land on their number: a part can sit under a tag we don't read, or be counted in two places. Treat a calculated number as our reading of the filing, not as the company's own figure. That is why each one is marked.

Known limits

• A fill-in can differ from the company's own figure. Short-term debt and long-term debt are the lines where our formula most often differs from what a company reports when it reports both. Some of that is companies using the same tag for different things, which no formula can settle. Where a cause is clear we fix it, and each fix is checked against the nightly comparison.
• A sum can read low without warning. The fill-in sums add whichever parts are filed under the tags we read. If a company moves a part to a tag we don't read, the total drops. We fix these as they are found, and the fix is made for every company with the same pattern, not for one. Total debt and liquid cash are read from the balance sheet, which doesn't depend on tags, and total debt has a check as well.
• Debt the balance sheet doesn't show as debt isn't counted. Some companies put the part of their debt due within a year inside a broader line such as "accrued expenses and other", and give the figure only in the notes. Total debt is what the balance sheet shows, so it reads lower than the company's own total in the notes.
• A finance arm's debt is the company's debt. A car maker's balance sheet includes the money its finance arm borrows to lend to customers, and a conglomerate's includes its railroad's and utilities' bonds. Total debt and net cash show the whole company, as its balance sheet does.
• No balance sheet, no number. Total debt, liquid cash and net cash start in 2009, when the SEC's balance-sheet data starts, and a newly listed company has none until the SEC's monthly data includes its first report.
• The newest quarter can wait a few weeks. Where a company's newest report can't be read with its last balance sheet's lines (a line on its own tag, a new line, a balance sheet shown per business), total debt or liquid cash for that quarter shows "no data" until the SEC's monthly data has it.
• Banks, insurers and property companies report differently. Some lines will be blank for them. A bank shows no total debt or net cash at all. For an insurer or a property company they are blank wherever the debt fails the check above.
• Who counts as a bank is decided by a rule, not by hand. A very small bank, with less than $100 million of deposits, isn't caught; its net cash, where shown, means little. We check the list the rule makes after each nightly update.
• A filed figure next to a calculated one. A company can file a line in one year and not the next. The filed year shows the company's figure and the others show ours, and the two can be built differently. A warehouse club that reports gross profit on merchandise sales alone will sit below our figure, which starts from all revenue. The line then steps where the solid stretch meets the dashed one.
• Capital spending is read from the cash flow statement only. A company that stops reporting it there shows "no data" for it and for free cash flow. It is what the company paid, before any money from selling equipment, so it can be higher than a company's own "net" figure.
• A short quarterly cash flow statement. Some companies show only the section totals in their quarterly reports, with no depreciation line. Their quarterly depreciation & amortization, and owner earnings, show "no data"; the yearly figures are read from the annual report.
• Revenue a bank doesn't total. Many banks show net interest income and non-interest income but no revenue total. Their revenue is the two added up, marked calculated. A lender that shows no non-interest income total, and a fund or trust, has no revenue line to read: "no data".
• A cash flow statement laid out out of order. Each section runs from the total before it to its own total. In the rare filing whose lines are stored with a section's total far from its lines (4 of about 5,000 cash flow statements in two months of SEC data), that section's lines aren't found, and show "no data".
• Statements shown per legal entity. A company whose statements are tagged only per subsidiary, with no company-wide figure, shows "no data" for the lines that are tagged that way.
• The newest year can wait a few weeks where a company changed a line's tag in its newest report: that report is read with the last statement's tags until the SEC's monthly data has it.
• Owner earnings treats all capital spending as upkeep. That is cautious on purpose, and reads low for a company that is spending to grow.
• Net income worked out from other figures. Where a company files no net income of its own, we work it out from its consolidated figure or from the figure available to common shareholders. A minority holders' share or preferred dividends that the company doesn't file count as none, because the accounting rules make a company that has them show them. A company that filed them in an unusual way, or has other adjustments between those lines, would read a little off. Where companies file both, the nightly comparison checks our working against their own figure.
• A company's own mistakes in its filing (a wrong date, a wrong tag) are shown as filed. We don't correct them.
• Numbers can differ from other sites. Other sources may use adjusted figures, a different choice of tags, or official closing prices.

Changes to how we calculate

When we change a formula, or stop using one, it is listed here with the reason.

October 11, 2026

After the first night of reading the statements we checked 32 more companies' filings. Each fix is one rule for every company.

• Bank revenue. A bank that prints no revenue total now has its net interest income and non-interest income added up, marked calculated, as banks that do print one count it. Before this, about 200 banks showed "no data", and a bank showing a fee line under a customer-revenue tag showed that fee line as its revenue.
• Depreciation & amortization is found in a direct-method cash flow statement's reconciliation below the statement, and under every standard tag that names depreciation or depletion (a utility's, a paper maker's). A company that owns equipment and rents it out has that depreciation counted; a tenant's leases are still left out.
• Capital spending is read from more of companies' own wordings ("invested in capital projects", "property and equipment additions", "improvements to operating real estate"), and from a line shown "net of" grants. Buying businesses, even under an equipment tag, and a bank's spending on foreclosed property no longer count.
• Revenue includes the SEC's total-revenue tag for companies that show their interest and dividend income separately.
• A table filed in a statement's place is skipped. Now and then the SEC's data has a note's table where a company's income statement should be. That filing is skipped for that statement, and its year is read the way the newest report is. So is a year whose annual report the SEC's monthly data lacks.
• Net income tagged as continuing operations. More and more companies with nothing discontinued tag their "Net income" line as income from continuing operations, which it then is. That line is now read, as is a company's own line worded plainly "Net income". Some companies had lost their place in the screener over this.
• Net income from the figure available to common shareholders adds back preferred dividends only. Since October 10 the minority holders' share had been added as well, which gave the consolidated figure for the few companies read this way.
• A revenue total "net of interest expense" (a card company's) is read as the total it is. On October 11, for a few hours, the words after "net of" were dropped for revenue lines too, and a card company's revenue after credit losses was read as its revenue.

October 10, 2026

• Revenue, net income, operating cash flow, capital spending, depreciation & amortization and dividends are read from the company's own income and cash flow statements, not from a list of tags. A company's capital spending under a tag the list didn't have now shows (a phone company's, for 2019 onward; a chip maker's own tag, for 2017 to 2021), and so does depreciation filed under a company's own tag. Capital spending and dividends are the amounts paid, even where a company's tags carried the opposite sign (one property company's dividends had shown as negative). Revenue is the statement's own total, not the largest figure under any revenue tag. How.
• What counts: capital spending includes software and a property trust's improvements to its buildings, not its purchases of buildings; depreciation & amortization adds up the lines a company shows separately; dividends include preferred dividends and leave out payouts to minority owners of subsidiaries.
• Another company's statements are never read as the company's own. A filing that is not an annual or quarterly report is read only for a period none of those shows. Before this, an 8-K carrying the statements of a business a company had bought could win by being filed later. This applies to the balance sheet too.

October 9, 2026

We checked every number the site shows against one rule: it is as filed, worked out exactly from what the company filed for that same period, or "no data". We ran the calculations on the full filings of 39 very different companies. These changes came out of that.

• Numbers are read from financial reports only, never from proxy statements. A proxy statement's executive-pay table repeats five years of net income. Being filed later than the annual report, it was taken as a restatement, and one company's table, entered a thousand times too small, showed its net income as about zero for five years.
• Net income is the company's own, never a stand-in. Where a company doesn't file its net income as one figure, we work it out from what it did file: net income including minority holders' share, minus that share; or, where that isn't filed either, net income available to common shareholders plus preferred dividends. A company with minority holders has to show their share in its income statement, so a period with no such figure has none. Before this, the figure including minority holders' share stood in for it, which for one broker was four times the company's own, and in other cases net income after preferred dividends did.
• A part filed in other periods but not this one is unknown, not zero. A sum such as short-term debt or tangible equity now shows "no data" for a period where the company left out a part it reports in other periods. A part a company never files still counts as nothing (it doesn't have one).
• Depreciation & amortization is shown only where the company files it as one figure. Before this, we added up the two parts we could read, which left out other write-downs and read low. Owner earnings is blank where it is.
• Dividends per share are no longer worked out by subtraction for a quarter. A special dividend paid in one quarter could land in another.
• Return on invested capital is blank for a year with no liquid cash figure. Before this, the cash counted as zero there.
• Total debt is what the balance sheet shows. A company's own figure for all its debt, from its notes, no longer replaces it when larger.
• The newest report, read with the last balance sheet's lines, is used only where every total on it still adds up, not just the four main ones.
• The screener never carries a figure over from an older year: each column is the latest fiscal year's own figure, or empty. Earnings per share and dividends per share in the screener are the last annual report's, as filed, the same figures the chart's P/E and dividend yield use. Earnings per share is no longer estimated from a year's profit and one quarter's share count.
• The preview picture of a shared link draws a year with no number dotted, as the chart does, instead of solid, and names a line with no number for the newest year "no data" instead of showing an older year's figure.
• On a yearly chart, each year's label sits under that year's point. Before this, the labels marked January 1, so a year ending December 31 sat beside the next year's label.
• A line you tick that has no number in the years shown now says "no data" in its label above the chart, instead of drawing nothing without a word.
• Fiscal years are named the way companies name them. A year ending in the first days of January is named for the year before, so each year has its own place on the year slider.

October 8, 2026

• Total debt and liquid cash are read from the balance sheet as filed, and only from it, not guessed from a list of tags. Where there is no balance sheet to read, they show "no data". The thirty-odd extra debt labels tried since October 7 are no longer used for total debt. Debt on a company's own tag is now counted (a property trust's credit line), as are Treasury bills a company holds as short-term investments and marketable securities listed in current assets. Restricted cash is left out. A company whose balance sheet shows debt only per business now has a total debt figure. How.
• Every sum we add up is checked against the company's own total. Liquid cash, short-term debt, long-term debt, operating lease liabilities and gross profit are left blank on a date where they come out bigger than the total they are part of, and so is everything built on them. How.
• Banks no longer show total debt or net cash, or the lines built on them. A bank's cash is mostly its customers' deposits, so net cash said little and could read as a large number either way. Before this, most banks were blank anyway, because the interest they pay on deposits failed the debt check, but a few showed a figure. Now every bank is treated the same. Who counts as a bank.

October 7, 2026

We ran the calculations on the full filings of twelve very different companies and compared the results with their annual reports. These changes came out of that.

• Total debt is checked against the interest bill, and left blank, with net cash, where the check fails. Before this, a company whose debt sat under tags we don't read could show almost no debt and a net cash figure that was really just its cash.
• Same evening, after the check's first run across every company: it left about one company in five blank, too many. So before a year is left blank, thirty-odd other debt labels are now tried (largest wins). The "small beside the cash" limit went from 0.5% to 2%. And total debt can no longer be more than total liabilities.
• Two more debt tags are read: the single long-term debt figure that banks and others file when their balance sheet isn't split by due date, and a company's own figure for all its debt, used when it is larger than our sum.
• Commercial paper is no longer counted twice. It is one part of short-term borrowings, and where a company filed both, short-term debt and total debt read too high. One of the two is now read, never both.
• Per-share figures and share counts are split-adjusted on every chart. Before this they were adjusted only where share prices were loaded. On other charts a split showed as a collapse in earnings per share.
• "Annual" means fiscal years only. A "twelve months ended" column in a quarterly report is no longer charted as a year.
• One more tag is read for short-term investments: the one most companies used before 2018. Liquid cash and net cash read too low in those years.
• One more tag is read for revenue: a bank's total net revenue, which some banks file every quarter where the usual tag appears only once a year.
• The figure beside each line, and each tile, is now the newest period or a dash. Before this, a line the company stopped reporting kept showing its last figure, with no date.

October 5, 2026

• Calculated numbers are now drawn as a dashed line. Before this they were named in a note under the charts. The line itself now says it.
• Total liabilities, when a company doesn't file it, now also takes off temporary equity: shares and minority interests the holder can hand back for cash. A balance sheet lists these between liabilities and equity. Before this change the figure read high for companies that have them.
• A result below zero is left blank for long-term debt and total liabilities. Neither can be negative, so a negative result means the parts weren't what the formula assumed.
• One more tag is read for short-term investments, used only when a company files none of the others.
• One more tag is read for long-term investments, also used only when a company files none of the others. Some companies used it in earlier years, and those years showed "no data".

Every line, and how we get it

In the same order as the app. As filed is read straight from the filing. As filed, calculated when missing uses the formula only when the company filed no figure. Calculated is always our arithmetic. Parts in italics count as nothing only for a company that never files them; one filed in other periods but not this one leaves the sum blank.