SEC Filings and Earnings Calls Are the Source. Here's How We Turn Them Into Signals

Every signal on this platform traces back to a primary source document: an SEC filing, an earnings call transcript, or an investor presentation filed with a regulator. That grounding is deliberate. When a management team says something on the record, they are legally accountable for it. When a company discloses a risk factor, a supplier dependency, or a new capital commitment in an SEC document, that disclosure carries weight that no analyst summary or press release can replicate. SeventhBiz Intelligence is built on that foundation, and understanding the methodology helps you use the platform with more precision.

The Source Layer: Primary Documents Only

The research pipeline ingests filings directly from EDGAR and official transcript providers. 10-Qs, 10-Ks, 8-Ks, S-1s, proxy statements, and earnings call transcripts form the raw material. Nothing enters the signal set that cannot be attributed to one of those sources.

That discipline matters in practice. Take the biopharma outsourcing shift currently running through the sector. The finding that 24 companies in the tracked universe explicitly disclose third-party CMO/CRO dependencies comes from reading those disclosures, not from inferring them from press coverage. When SANA's decision to exit internal cell therapy manufacturing in favor of CDMOs is flagged as a threshold signal, that call is grounded in what SANA actually filed, not in what an analyst predicted they might do. The $800 million AZN deployed across Dizal, Sino Bio, and CSPC AI is a disclosed figure, not an estimate.

Verbatim management quotes sit inside every brief for the same reason. When Fluence Energy disclosed its first large behind-the-meter order and hyperscaler awards, the filing language is preserved exactly:

"the Company's first large, behind-the-meter order signed during the third quarter and approximately $550.0 million of awards from a hyperscaler in July 2026" (FLNC)

That quote does not get paraphrased into a softer, hedged version. It stays verbatim because the specific language, the dollar figure, the timing, and the customer category are all analytically meaningful.

The Analytical Layer: From Disclosure to Signal

Raw filings are not intelligence. The analytical layer is where SeventhBiz adds the interpretive work that transforms a disclosure into a signal a deal team can act on.

Signals are classified by type: NEW (first-time disclosure of a dynamic), RISING (an escalating trend with directional evidence), THRESHOLD (a quantitative or qualitative inflection that changes the investment calculus), and RISK (a disclosed exposure with forward earnings or operational implications). These classifications are not editorial opinions; they are conclusions drawn from comparing current disclosures against prior periods and sector-wide patterns.

Consider how this plays out in the Bring Your Own Power space. Fluence's disclosure of master supply agreements with two hyperscalers is classified as a THRESHOLD signal, not merely a NEW one, because it represents a channel shift from utility-facing sales to direct hyperscaler relationships, a structural change with margin and revenue concentration implications. Bloom Energy's language shift, from customers exploring on-site solutions to customers actively deploying them, is tracked across multiple periods to establish directionality. The quote that captures it:

"customers are increasingly turning to on-site and islanded power solutions" (BE)

That language shift from one quarter to the next is the signal. The filing is the evidence.

The same logic applies to structural signals. When Exelon discloses regulatory proposals around bring-your-own-generation data center policies:

"Introduced proposals on transmission siting reforms and data center policies on bring your own generation and environmental reporting regulations" (EXC)

that disclosure is cross-referenced against hyperscaler capex disclosures, CDMO capacity announcements, and grid interconnection data to build a sector-wide picture, not treated as an isolated data point.

What AI Does and What It Does Not Do

AI accelerates the ingestion, classification, and pattern-matching work across thousands of filings simultaneously. A human analyst reading every 10-Q across the energy, biopharma, and data center sectors in a single quarter would take weeks. The platform processes that volume continuously and flags signals as filings drop.

What AI does not do is substitute for the primary source. Every quote is pulled verbatim. Every figure is disclosed, not modeled. Every signal classification is anchored to a specific filing event. The AI layer handles scale; the SEC filing handles credibility.

This matters most when a signal requires negative inference. When a major pharma company does not disclose a CDMO dependency in a cycle where 24 peers do, that absence is analytically meaningful, but it is treated as an open question requiring monitoring, not as a concluded finding. Absence of disclosure is not a fabricated signal.

How to Use This in Practice

For PE and VC teams evaluating biopharma outsourcing dynamics, the signal set tells you which large-cap buyers have explicitly disclosed inorganic pipeline strategies and which CDMOs are named as dependencies, directly from the filings that create legal accountability. For corporate development teams tracking power infrastructure deals, the verbatim disclosures from FLNC, BE, EXC, and ENPH give you the exact language management used when committing to hyperscaler relationships, which is the language that matters in diligence and negotiation.

The methodology is straightforward: primary sources first, analytical classification second, AI scale third. The result is intelligence you can trace back to a document number on EDGAR, which is exactly the level of accountability institutional deal work requires.