The Buying Signal Had a Date. The List Threw It Away.

A vendor sells you a list of 4,000 companies. The list is accurate. The contacts are deliverable. The industry codes are correct.

And it is worth almost nothing, because it does not tell you what happened, when it happened, or where to check.

The problem is not data quality in the usual sense. It is that a buying signal is an event and a list is a snapshot, and the snapshot has thrown away the only part that could justify a call.

What this analysis can and cannot support

Three limits belong at the top.

First, the regulatory and research material cited here comes from primary sources read in their published form: two SEC help pages, a Securities and Exchange Commission search interface, and two OpenAI posts dated November 10, 2026. Where a vendor describes its own product, that is labeled.

Second, no conversion rate, reply rate, lead-quality figure or customer result is asserted anywhere in this article. Nothing here establishes that any organization won a customer.

Third, the five-field record, the options, the questions, the test and the indicators are this publication's analysis. They have not been validated as a standard.

The difference between a name and an event

Consider what a salesperson actually needs in order to write a first sentence that is not embarrassing.

They need a specific thing that changed, recently, at a specific company, that a reasonable person could connect to a business need. That is an event. It has a date. It was published somewhere.

A list record usually contains a company name, an industry code, a headcount band and a title. All of those are stable. Stability is exactly what makes them unusable as a reason to call.

This is where the language gets sloppy in a way that costs money. Vendors describe purchased lists as containing buying signals. What most contain is firmographic fit — a description of the kind of company that might need you, not evidence that this one does right now.

The verification instrument has blind spots

Suppose you decide to check a claimed signal at the source. For a public company, that source is likely the Securities and Exchange Commission's EDGAR system, and EDGAR is candid about its own limits in ways that reward reading.

The Commission's Full-Text Search FAQ states that full-text search covers filings submitted electronically in the last four years. It also states something more consequential: the search engine is conceptual rather than keyword-based, and may retrieve results that are conceptually related to your search terms but do not contain your precise term.

Read that twice. It means a zero-result search is weaker evidence than it appears, and a hit is not proof the phrase is there. The FAQ suggests opening the document and using the browser's find function to locate the term inside it.

That is a small paragraph of search documentation, and it invalidates a habit many revenue teams have adopted: typing a company name into a filings search, seeing nothing, and marking the signal unverified.

A discrepancy worth flagging rather than resolving

Two SEC pages, both retrieved for this edition on November 10, 2026, describe coverage differently.

The Full-Text Search FAQ says the last four years. The advanced search interface at sec.gov/edgar/search states that it gives access to the full text of electronic filings since 2001.

This analysis does not resolve that difference, and a reader should not treat either number as settled. The practical lesson is procedural: check the current stated scope on the page you are actually using, because a verification window that quietly changes is worse than one you know is narrow.

The Commission's own guide to using EDGAR supplies the rest of the toolkit, and it is more useful than it sounds for qualification work. It describes searching by company name or CIK, limiting results by date or form type, and a Current Events Analysis view that retrieves filings made during the previous week.

A weekly window plus a date filter is, functionally, a freshness mechanism. Most purchased lists have no equivalent.

Meanwhile the place intent shows up is moving

Filings are one signal surface. Two announcements on November 10, 2026 point at a different one.

OpenAI introduced Businesses in ChatGPT, described in its own summary as a way to manage business information, get insight into how a business shows up, and turn interest into action inside ChatGPT.

The same day, OpenAI published research with Upwork finding shifting demand for freelance work, new skill combinations, and larger budgets and more hours for AI-related projects.

Both are described here only as the publisher described them. Neither is evidence about your market.

But the structural implication is worth stating plainly: if buyers increasingly ask an assistant rather than a search engine, then some purchase intent is being expressed in a surface no CRM is watching. That is not a reason to abandon filings. It is a reason to stop pretending that one signal surface is the whole market.

What a qualified record actually has to carry

Five fields, and the absence of any one of them is the difference between a lead and a name.

• What happened. A specific event, described in a sentence a stranger could verify.
• When it happened. Not when the list was compiled. The event date, distinguished from the publication date.
• Where it was published. A URL, ideally primary, that a reviewer can open without an account.
• Who owns the reason now. The role or person for whom the event creates a problem, checked for whether they still hold that role.
• When it expires. A date after which nobody is allowed to reference the signal again without re-checking it.

The fifth field is the one nobody includes and the one that prevents the most embarrassment. A three-year-old expansion announcement cited as current news is worse than no signal at all, because it damages the credibility of everything else in the message.

Three ways to source signals, and what each costs

Buy the list. Fastest, cheapest per record, and the weakest on the five fields above, because the compiler's commercial incentive is coverage rather than provenance.

Build from primary sources. Slower, and it produces fewer records. It also produces records whose third field is a URL you control the interpretation of. This is the option that survives a procurement review.

Combine a signal feed with internal evidence. The most durable, and the most work. A public event plus something only you can see — a renewal date, a support ticket pattern, a hiring shift in a role you sell to — produces a reason to call that competitors cannot copy from the same feed.

DealFORGE, per its own approved product description, sits in the second category: it finds current B2B lead opportunities using buying signals from official filings, hiring activity and predictive business needs, and it positions those as evidence-based opportunities to research and contact. That is a sourcing claim, not an outcome claim, and it should be read as one.

Four questions before you buy another list

• What is the event behind each record, and is it dated? If the answer is a firmographic attribute rather than an event, you are buying a filter, not a signal.
• Can I open the source myself? A record whose provenance is the vendor's own database cannot be defended to a customer, an auditor or your own sales leader.
• How old is the oldest record you will accept? Set the number before the file arrives, not after.
• Who checks that the reason still exists? Signals expire quietly. Somebody has to own the expiry.

A 30-day test you can run on one segment

• Days 1 to 5. Pick one segment and one signal type. Write down the event definition in a sentence, including what would not count.
• Days 6 to 12. Pull twenty records. For each, record the five fields, and mark every one where the source URL leads to a page you cannot open without an account.
• Days 13 to 18. Check ten of them against the primary source. Open the document rather than trusting the search result, for the reason the SEC's own FAQ gives.
• Days 19 to 25. Write the first sentence you would send for five of them, and have someone outside the team judge whether it reads as informed or generic.
• Days 26 to 30. Count how many records survived all five fields. That number, not the file size, is what you actually bought.

Four indicators worth baselining

The share of records carrying a verifiable primary-source URL. The median age of the underlying event, measured separately from the list's compilation date. The share of records where the named role still exists at the company. And the number of minutes it takes a reviewer to reconstruct one record from scratch.

None requires a vendor benchmark, and each describes your pipeline rather than someone else's. Baseline them before the test. If a supplier cannot improve the first indicator, the other three will not move either.

What remains unestablished

• Whether attributable, dated signals produce better outcomes than firmographic lists at any organization. Nothing here establishes a result.
• What DealFORGE's lead quality, conversion rate or accuracy is in any customer's use. No such figure is asserted.
• Whether intent expressed inside AI assistants is measurable, addressable, or durable as a signal surface.
• What the correct coverage window for EDGAR full-text search actually is, given the conflicting statements noted above.
• Whether the OpenAI and Upwork research on freelance demand generalizes beyond the populations it studied. The methodology was not reviewed for this edition.
• Market data. None was reviewed, so no stock movements are reported.

Evidence notes and limits

The SEC's Full-Text Search FAQ is cited for the statement that full-text search covers filings submitted electronically in the last four years, for the description of the engine as conceptual rather than keyword-based with the caution that results may be conceptually related without containing the precise term, and for the guidance to open a document and search within it. The advanced search interface at sec.gov/edgar/search is cited for its statement that it provides access to the full text of electronic filings since 2001; the difference between that figure and the FAQ's four years is reported here as a discrepancy, not resolved. The Commission's guide, How Do I Use EDGAR?, is cited for company-name and CIK search, date and form-type limiting, the Current Events Analysis view covering the previous week, and its pointer to full-text search of the last four years. All are U.S. government publications describing their own systems.

OpenAI's Businesses in ChatGPT post and its research post with Upwork are both dated November 10, 2026 and are cited only for how each is described in its published summary: managing business information and turning interest into action inside ChatGPT, and research finding shifting freelance demand, new skill combinations and larger budgets and more hours for AI-related projects. Neither methodology was reviewed, and no figure is drawn from either.

DealFORGE is described only as its owner-approved product description states: it finds current B2B lead opportunities using buying signals from official filings, hiring activity and predictive business needs, and those are positioned as evidence-based opportunities to research and contact rather than guaranteed purchases or closed accounts. The free-trial terms appear only in the separate advertisement.

No pricing page, contract, independent benchmark, third-party audit or customer case study was reviewed for this edition. The reading of signals as dated events, the five-field record, the three sourcing options, the four questions, the 30-day test and the four indicators are this publication's analysis. Nothing here establishes that any organization improved win rates, pipeline or staff capacity.

The advertisement above is separate from this analysis and is not a source for any claim in it.