Published: October 29, 2026

By Nir Dukas, founder of SeeNote.

How many footnote cross-references are in a typical 10-K?

A typical large-company 10-K contains about 50 numbered footnote cross-references. In the latest 10-Ks of 28 Dow Jones Industrial Average companies we counted a median of 51.5 "Note N" references per filing and a median of 19 notes. The range runs from 7 at Apple to 347 at JPMorgan Chase, 1,940 in total.

This is our own count, made from the filings themselves. Every per-company figure on this page comes from the published CSV, and the method and its limits are set out below so you can check or extend it.

What are the headline numbers?

We took the latest 10-K of each of the 30 companies in the Dow Jones Industrial Average as it was constituted after Nvidia and Sherwin-Williams joined in November 2024, and counted every numbered reference to a note: "see Note 12", "Refer to Note 17.", "Notes 5 and 7". Two companies could not be measured this way (explained below), which leaves 28 filings with fiscal years ending between September 27, 2025 and July 25, 2026.

• 28 companies analysed, 551 numbered notes, 1,940 numbered references in total.
• Median 19 notes and 51.5 numbered references per 10-K. The mean is 69.3, pulled up by two banks.
• Range: 7 references at Apple to 347 at JPMorgan Chase.
• Median density: 6.9 references per 10,000 words, from 1.9 at Salesforce to 18.1 at JPMorgan Chase.
• 803 references, or 41.4 percent, sit inside the notes themselves and point from one note to a different note. The rest are in MD&A, the face statements, the auditor's report and other parts of the filing, or are a note citing itself.
• 243 of the 551 notes (44 percent) point to at least one other note.

The median is the better guide to a typical filing. Fourteen of these 10-Ks carry 52 or more numbered references and fourteen carry 51 or fewer, while the mean of 69.3 mostly reflects JPMorgan Chase and Goldman Sachs.

How many cross-references does each Dow company's 10-K contain?

The table is sorted by the number of numbered references. Density is references per 10,000 words of the whole 10-K, which lets you compare a 32,521-word filing (Apple) with a 192,212-word one (JPMorgan Chase).

Source: our count of each company's latest 10-K on EDGAR (company names link to the filing). Words are counted in the 10-K's main document as plain text. Full data, with filing dates and more columns: cross-references-in-10-ks.csv.

Why do bank 10-Ks have so many more references?

JPMorgan Chase and Goldman Sachs account for 549 of the 1,940 references, 28.3 percent of the total, from 64 of the 551 notes. Their 10-Ks are the two longest in the group (192,212 and 177,256 words) and they have the most notes (34 and 30). Length explains part of it. Density explains the rest: JPMorgan Chase runs at 18.1 references per 10,000 words, the highest of the 28, and Goldman Sachs at 11.4.

Bank notes depend on each other more than most. A loans note leans on the allowance for credit losses, the fair value notes lean on the derivatives and securities notes, and variable interest entities connect to securitizations and to the consolidation policy. At JPMorgan Chase, 28 of 34 notes point to another note, and at Goldman Sachs 23 of 30, against 44 percent across all 28 companies. At JPMorgan Chase, 153 of the 347 references sit inside the notes; at Goldman Sachs, 138 of 202. The single most-referenced note in the whole data set is JPMorgan Chase's Note 6, Noninterest revenue and noninterest expense, with 42 references.

Size alone does not produce references, though. Travelers, an insurer, files a 126,844-word 10-K, the third longest here, yet it has 51 references and a density of 4.0. The next three densest filings after JPMorgan Chase are Sherwin-Williams at 16.1, Johnson & Johnson at 14.9 and Visa at 14.7 references per 10,000 words. Coca-Cola, at 13.5, often gives a reference a sentence of its own: the bare form "Refer to Note N." appears 33 times, for example "Refer to Note 17." If you want background on why the documents themselves keep growing, see why 10-Ks keep getting longer.

Why does Apple's 10-K have only seven?

Apple's 10-K for the fiscal year ended September 27, 2025 is the shortest in the group at 32,521 words, with 13 notes. We found 7 numbered references, 5 of them to Note 7, Income Taxes. One of those five is in the auditor's report, which describes a critical audit matter "As discussed in Note 7 to the financial statements". Only 2 of Apple's 13 notes point to another note.

Apple's notes rarely cite each other: 3 of the 7 references sit inside the notes. When Apple does cite a note it tends to give the title as well, as in MD&A's "refer to Note 7, "Income Taxes" in the Notes to Consolidated Financial Statements". Salesforce is the other low outlier: 12 references in 63,421 words, a density of 1.9, the lowest of the 28. Cisco (3.8), Microsoft (4.4) and Nvidia (4.5) are also near the bottom, and by this count Nvidia's 17 notes never cite each other at all.

Why are McDonald's and IBM left out?

McDonald's does not number its notes. It refers to them by title and page: its 10-K says "Refer to the Income Taxes footnote on page 58 of this Form 10-K for additional information." A count of "Note N" patterns would score McDonald's at zero, which would misstate how much cross-referencing its filing contains, so we excluded it rather than report a false zero.

IBM's financial statements are not in the body of its 10-K. The 10-K incorporates them by reference from IBM's 2025 Annual Report to Stockholders, and IBM letters its notes rather than numbering them: "Refer to note Q, "Commitments & Contingencies," on pages 91 to 93 of IBM's 2025 Annual Report to Stockholders". Neither filing fits a numbered-note count.

How did we count?

1. Take each company's most recent 10-K from EDGAR and convert the main document to plain text.
2. Find the note headings (Note 1, Note 2 and so on) and keep the longest run of headings that counts up from 1, which marks where each note begins and ends.
3. Match "Note N" and "Notes N and M" anywhere in the document, case-insensitive, with lists joined by commas, "and", "through" or "to".
4. Drop matches on the note heading lines themselves and on contents-list lines (a line that holds only a note label and title, with or without a page number, or a bare "Note N" label). Also drop matches whose number exceeds the number of notes, and matches preceded by a debt word such as "senior", "convertible" or "floating", so that debt instruments called notes are not counted.
5. Record which note each reference sits in, if any, to separate references inside the notes from those in MD&A and elsewhere.
6. Divide the reference count by the document's word count and multiply by 10,000.

This is version 2 of the count. The first version also counted contents-list lines and missed Goldman Sachs's Note 30; both are fixed in the figures and CSV here. The code is short and deterministic; anyone with the same filings and the same rules gets the same numbers. Per-company detail, including the most-referenced note at each company and the chain measures used in our later data pages, is in the CSV.

What are the limitations?

Read the figures as a count of a specific pattern, not of every pointer a reader might follow.

• Title-only references such as "see Income Taxes" or "the Income Taxes footnote" are not counted, so filings that name notes by title look sparser than they read.
• A range such as "Notes 1 through 5" counts as two references (the two endpoints), not five.
• The count covers the whole 10-K, so it includes references in MD&A, the face statements and the auditor's report as well as in the notes.
• Tables inside a note that tell you which note holds each accounting policy count as references, one per row. JPMorgan Chase's Note 1 and Goldman Sachs's Note 3 both contain one.
• The heading test is a pattern, not a reader. A contents line is skipped only if it consists of a note label and a title (with or without a page number) or a bare "Note N" label, so an index laid out differently would still be counted.
• Twenty-eight filings from one index is a sample of very large US companies. Smaller filers, other industries and 10-Qs may look different.

How should you cite this data?

The data is free to reuse under the Creative Commons Attribution 4.0 license. Please credit SeeNote and link to this page. Suggested citation:

SeeNote (2026). Numbered footnote cross-references in the latest 10-Ks of 28 Dow Jones Industrial Average companies. https://seenote.co/data/cross-references-per-10k

Download the data: cross-references-in-10-ks.csv (28 rows, one per company; columns for ticker, company, fiscal period end, filing date, EDGAR URL, notes, numbered references, references inside notes, document words, references per 10,000 words, most-referenced note and its count, notes linking out, longest chain in hops and reachable note pairs).

For a reader's view of what these references point to, see how to read 10-K footnotes.

Questions

Following dozens of references without losing your place

A median of 51.5 numbered references means a typical 10-K asks you to jump to another part of the document dozens of times, and EDGAR leaves each jump to scrolling or Ctrl+F. SeeNote is a Chrome extension that turns those "see Note 12" references into links that open the referenced note verbatim in a popup where you are reading, on EDGAR 10-Ks and 10-Qs. The first three filings are free, with no card and no expiry; pricing is on seenote.co.