When A Headline Names One Company, EDGAR Full-Text Search Names The Rest

A headline names one company. The position that follows usually sits in that same one company, because that is the name on the screen. But what the headline describes is almost always a relationship — a plant that stopped, an order that was pulled, a contract that was signed, a certification that was withdrawn. Relationships have two ends. The second end is rarely in the headline.

The second end is not a matter of inference. For companies that report to the U.S. Securities and Exchange Commission, dependence is written down: in a numbered section of a specific form, in language the disclosure rules require, in a database that has been text-indexed since 2001. The work is not guessing who is attached to whom. The work is knowing which section holds that sentence and how to search for it backwards.

The Rules Say Where Dependence Must Be Disclosed, Not How Much Of It Is Too Much

Form 10-K is a shell. Nearly every substantive item in it points at a section of Regulation S-K, codified at 17 CFR Part 229. Item 1, Business, points at 229.101. Item 1A, Risk Factors, points at 229.105. Item 2, Properties, points at 229.102. Item 15 points at 229.601, the exhibit list. Reading a 10-K without knowing that mapping is reading a table of contents.

Item 101(c) is where dependence lives. Its subparagraph (c)(1)(i) calls for disclosure of “Revenue-generating activities, products and/or services, and any dependence on revenue-generating activities, key products, services, product families or customers, including governmental customers.” Subparagraph (c)(1)(iii)(A) covers “Sources and availability of raw materials.” One clause faces downstream toward buyers, the other faces upstream toward inputs.

Two things about that text matter more than the text itself.

• There is no percentage threshold in it. Item 101 does not say ten percent, or twenty, or any number. The SEC's 2020 adopting release, Release No. 33-10825, effective 9 November 2020, describes the amended item as expanding a principles-based approach with “a nonexclusive list of disclosure topic examples.” The operative test in the rule text is materiality, and materiality is the filer's judgement.
• Disclosure is therefore a statement about the filer's own view. A company that spells out a concentration in Item 1 has decided that concentration is material to an understanding of its business. That decision is information, separate from the number attached to it.

Item 105 is the same fact restated as a hazard. Paragraph (a) requires material risk factors under a Risk Factors caption, each with its own subcaption, and pushes risks that “could apply generically to any registrant or any offering” to the end under a General Risk Factors heading. Paragraph (b) adds a length mechanic worth knowing: if the risk factor discussion runs longer than 15 pages, the filer must put a bulleted or numbered summary of no more than two pages in the forepart. When that two-page summary exists, it is a filer-written ranking of its own risks, and a concentration that survives into it has been ranked near the top by the people who know the business.

Item 102 is the least-read of the four and often the most concrete. It calls for disclosure, to the extent material, of “the location and general character of the registrant's principal physical properties,” and its Instruction 1 asks for the “suitability, adequacy, productive capacity, and extent of utilization” of those properties. That is a list of where things are physically made, and roughly how full the lines are. A weather event, a port closure or a regional power problem is an Item 2 question before it is anything else.

The Counterparty Is A Signatory Before It Is An Inference

Narrative disclosure describes a relationship. The exhibit list attaches it. Under 229.601(b)(10), material contracts are filed as exhibits. The general rule carries a wide carve-out — contracts that ordinarily accompany the kind of business the registrant conducts need not be filed — but the carve-out has exceptions, and one of them is built for exactly this problem.

Subparagraph (b)(10)(ii)(B) requires the filing of “Any contract upon which the registrant's business is substantially dependent, as in the case of continuing contracts to sell the major part of registrant's products or services,” along with major purchase requirements contracts, franchises and licenses material to operations. A supply relationship that the business genuinely leans on does not stay off the exhibit list merely because supply contracts are ordinary for that industry. The document goes in, with signature blocks, and a signature block contains a name.

The same paragraph contains one of the few hard numbers in this corner of the rules. Subparagraph (b)(10)(ii)(C) reaches contracts for the acquisition or sale of property, plant or equipment for consideration exceeding 15 percent of the registrant's fixed assets. Note what that threshold is about: assets, not customers. It is a common error to remember the existence of a number in Item 601 and then apply it to customer concentration in Item 101, where no number exists.

Filed does not mean fully legible. Subparagraph (b)(10)(iv) lets a registrant redact information that is both immaterial and of the type it treats as private or confidential, provided the redacted exhibit carries a prominent statement that portions have been omitted and marks the locations with brackets. In practice this means a filed supply agreement can name both parties and the term while the volumes, unit prices and rebate schedules sit behind brackets. The relationship is confirmed. The economics may not be.

The exhibit route also has a fast lane. A material definitive agreement entered into outside the ordinary course triggers Form 8-K Item 1.01, and its termination triggers Item 1.02. Form 8-K's General Instruction B.1 sets the deadline: “a report is to be filed or furnished within four business days after occurrence of the event.” A contract signed on a Monday is on the wire, frequently with the contract itself attached as Exhibit 10.1, before the following Monday.

Running The Search In The Other Direction

Everything above describes a company disclosing its own dependencies. The more useful move is the inverse: take the name in the headline and find every filing by somebody else that had a reason to write that name down.

EDGAR full-text search does this. Its published scope is broad in the way that matters here. The SEC's own FAQ for the tool states that it searches “the full text of all EDGAR filings submitted electronically since 2001,” and that “The full text of a filing includes all data in the filing itself as well as all attachments (such as exhibits) to the filing.” Contracts are attachments. Press releases filed as Exhibit 99 are attachments. The index reaches them.

The syntax is small enough to memorise. Quotation marks force an exact word sequence. OR and NOT work when capitalised, and a leading hyphen is shorthand for NOT. NEAR(n) requires two terms within n words of each other, so Alpha NEAR(5) supplier is a different and much sharper question than Alpha supplier. A trailing asterisk stems a word. Results filter by form type, by date range in yyyy-mm-dd, and by company name or CIK.

A concrete illustration of what the index reaches: the exact phrase "exclusive supply agreement", restricted to filings dated 1 January to 31 December 2025 with no form filter, returned 135 documents. The file types of the first eight hits were EX-10.2, EX-99.1, EX-99.1, 8-K, EX-10.1, EX-10.6, 10-K and 10-Q, attached to root forms S-1, 8-K and 10-K. Six of eight were exhibits rather than narrative body text. Searching only the readable part of the annual report would have missed most of them.

Direction has to be established separately from existence. A hit tells you two names appear in one document. It does not tell you which side depends on which. That answer sits in the Item 1 of each company in turn: what the filer says it depends on, and what the counterparty says in its own Item 1, are two independent statements that can be read against each other. When only one side discloses the relationship as material, the asymmetry is itself the finding.

The vocabulary companies use is narrow and repeats, which is what makes the reverse lookup practical. Counting Form 10-K filings dated in calendar 2025 that contain each exact phrase gives a sense of how standardised the language is.

The gap between the columns is the useful part. Supply-side language is roughly twice as common as customer-side language in this sample: 883 filings used “supply agreement” and 861 used “sole source”, against 416 for “customer concentration” and 203 for “our largest customer”. That is not evidence that customer concentration is rarer. It is evidence that upstream dependence gets named in standard phrases while downstream dependence gets described in prose that varies filer to filer. A search built only on customer-side phrases will systematically under-return, and the fix is NEAR() against a specific company name rather than a longer list of guessed phrases.

What These Documents Withhold, And How Late They Arrive

Four limits do most of the damage if they are not held in mind.

Latency. Form 10-K General Instruction A.(2) gives large accelerated filers 60 days after fiscal year end, accelerated filers 75 days, and all other registrants 90 days. An annual report is therefore at least two months stale the day it appears and up to a year stale before the next one lands. The four-business-day Form 8-K path is the only fast channel, and it fires only for events the rules enumerate.

Redaction. As above, 229.601(b)(10)(iv) permits omission of immaterial confidential terms. Names survive redaction far more often than prices do.

Strings, not entities. Full-text search matches characters. Legal names, trade names, abbreviations, subsidiary names and post-merger renamings are different strings for the same business, and an exact-phrase query catches one of them. Running the query several ways, and using NEAR() with a distinctive product or facility name, is the partial remedy.

Silence is not evidence. Because Item 101 sets no numeric trigger, a filing with no percentage in it may describe a company with no concentration or a company whose management concluded its concentration was immaterial. Those are different states of the world producing identical text. It is worth stating plainly what could not be verified for this piece: no percentage threshold for customer dependence appears in the text of Item 101, Item 102 or Item 105. Where a concentration percentage does appear in a filing, it arrives from somewhere other than these three items, and the first question is which section of the document it is sitting in.

What Would Invalidate This

This frame assumes the thing moving is a disclosed bilateral relationship. Several common situations break that assumption.

• The exposure is not bilateral. If a move is driven by an input price, a currency, a rate, or an index rebalance, there is no counterparty filing to find. Reading exhibits will produce a plausible story about a relationship that is not what is being priced.
• The counterparty does not file. Private companies, foreign private issuers on different forms, and non-listed subsidiaries are partly or wholly outside this index. Entire tiers of a supply chain can be invisible to EDGAR while being decisive to the outcome.
• Nothing here is private. These documents are public and machine-indexed. Any advantage comes from reading faster and more completely than the alternative, not from access. Treating a public filing as though it were a private edge is the most reliable way to be wrong at size.
• Sensitivity is not symmetric. A supplier deriving the majority of revenue from one buyer and that buyer sourcing a small fraction of its inputs from that supplier are the same relationship seen from two ends, with very different consequences at each end. Finding the link says nothing about how much either side moves.
• The disclosure may be obsolete. A dependence described in a filing made ten months ago may have been renegotiated, dual-sourced or terminated since. If it was terminated by a material definitive agreement, Item 1.02 should show it; if it was quietly diluted, nothing will.

And the standing condition: if a position is sized such that being wrong about direction is not survivable, the quality of the research on the second end of the relationship does not matter.

Concrete Framework

1. Write down the name, and what the headline claims happened to it. Separate the event from the entity before searching. The event determines which form to filter on.
2. Read the named company's own Item 1 first. Look specifically for the (c)(1)(i) material — dependence on products, services or customers — and the (c)(1)(iii)(A) material on sources and availability of raw materials. Note whether a percentage appears at all, and where.
3. Read Item 1A next, and check for the two-page summary. If the risk discussion exceeds 15 pages the summary is mandatory, and whatever the filer put in it is its own ranking of its own risks.
4. Read Item 2. Note the locations and the language about capacity and utilisation. Physical events resolve here.
5. Open the exhibit index and look for Exhibit 10 entries. A contract that appears despite being ordinary-course is there because of the substantially-dependent exception, which is a disclosure by placement alone.
6. Now run the reverse query. Put the company name in quotation marks in EDGAR full-text search, set the date range, and leave the form filter off for the first pass so exhibits are included.
7. Triage the hits by file type before reading any of them. EX-10.x is a contract. 8-K is a dated event with a four-business-day clock behind it. 10-K and 10-Q hits are narrative.
8. Re-run with NEAR() against the terms that matter. Pair the name with supplier, customer, terminate, or a specific product or facility name. Then re-run with alternate spellings, the former name, and known subsidiary names.
9. Establish direction explicitly, in writing. For each candidate counterparty, state which of the two disclosed the relationship as material, and which did not. Do not proceed on a link where neither side did.
10. Date every item you rely on. Attach the filing date and the fiscal period to each fact. A 90-day filer's annual report and a four-day 8-K are not the same kind of evidence and should not be weighted as though they were.
11. Write the disconfirming condition before sizing anything. Name the specific observation that would show the relationship is not what is being priced, and decide in advance what happens to the position when that observation appears.