Somewhere in a pipeline there is a lead scored 87. Ask what happened at that company, when it happened, and which document said so, and the answer is usually a vendor name.
That is not a data-quality complaint. It is a missing-field problem, and it costs hours every week: a representative who cannot see the event behind a score either researches the company again from scratch or sends the email and hopes.
This is also not a list-decay argument. Decay is about records going out of date. This is about records that never carried a date worth reading in the first place.
Three signal families, three different clocks
This section is analysis.
DealFORGE's approved product description positions it around buying signals from official filings, hiring activity and predictive business needs. Those are three genuinely different kinds of evidence, and the difference is not academic. Each one ages at a different rate and fails in a different direction.
A filing is a document. It was submitted on a known day, it carries an identifier, and its text does not change afterward. A job posting is an advertisement. It appears on a date, it can be withdrawn or reposted, and it describes an intention rather than a commitment. A predictive score is an output. It has a generation time but no publication date, and no external party can inspect the reasoning behind it unless someone wrote that reasoning down.
Blend all three into one number and the buyer loses the only thing that made the number actionable: which kind of evidence produced it.
What an official filing actually establishes
SEC EDGAR is where public companies submit disclosure documents, and each submission carries a filing date and an accession number that makes it individually retrievable. That structure is why filing-based signals are the most defensible of the three: the source is public, dated and quoted rather than summarized.
The limit is just as clear. A filing establishes that a company reported an event or a condition. It does not establish that the company is shopping for anything, has a budget, or would take a meeting.
A workforce reduction disclosed in a filing may indicate cost pressure, or it may indicate a completed restructuring. A new credit facility may indicate capacity to invest, or it may indicate a refinancing. The document is the observation. The commercial implication is the reader's inference, and those belong in different fields.
What hiring activity establishes, and the distinction that matters
This section is analysis.
Hiring is the most intuitive of the three signals and the easiest to misread. A cluster of postings looks like growth. It can also be churn, a reclassification, or a recruiter posting the same role in six cities to build a pipeline.
The Bureau of Labor Statistics draws a line worth borrowing. Its Job Openings and Labor Turnover Survey reports job openings, hires and separations as distinct measures rather than folding them into a single hiring number. An opening, a hire and a separation describe different states of the same position.
A job posting is none of those things precisely. It is an advertisement for a possible opening. That does not make postings useless as a signal; it makes them a signal that needs a date and a stated reading. "Forty postings in the last 30 days" is an observation. "This company is expanding its warehouse footprint" is an inference that may be wrong.
What a predictive score actually establishes
A model output is an estimate produced at a moment in time. It is useful for ordering a queue. It is not evidence, and it becomes misleading the moment it is stored next to a filed document without a label.
NIST's AI Risk Management Framework organizes its Core around four functions: Govern, Map, Measure and Manage. Read narrowly, that is a reminder that an estimate needs an owner, a stated purpose and a way to check whether it is working. The observation that scoring outputs are frequently stored with no owner attached is this publication's analysis, not a finding in the framework.
Why a blended score hides the only field a representative can use
This section is analysis.
A score of 87 that averages a filed document, a revocable posting and a model estimate is arithmetically fine and operationally useless. The representative cannot tell whether to open with a compliance question, a capacity question or nothing at all.
Worse, the three inputs decay at different speeds and the composite score does not say so. If the filing is nine months old and the posting is from last week, the number is dominated by whichever input the model happened to weight, not by what is current.
There is a second cost that rarely appears in a vendor comparison. A score with no visible evidence cannot be argued with. A representative who suspects a lead is wrong has no way to record why, so the same bad record returns next quarter with a fresh number attached.
The qualification record: five fields
Whatever tool produces the signal, the record a representative works from should carry five things.
- The event. What was observed, in one sentence, in plain language. "Filed a Form 8-K on a named date" is an event. "Growing fast" is not.
- The date. The date of the underlying publication, not the date the record entered your system. Those are different fields and should never be the same field.
- The source. A retrievable location. If nobody can reopen it, it is hearsay with good formatting.
- The inference. The possible commercial implication, explicitly labeled as an inference, with a confidence the author is willing to defend.
- The next question. What a representative would ask to confirm or reject the inference. This is the field that turns research into a conversation.
Five fields is not a heavy process. It is the difference between a record a colleague can pick up and a record only its author understands.
Three ways a team assembles signal evidence
- Manual research per account. Highest evidence quality, lowest volume, and the cost lands entirely on representative time. Workable for a named-account strategy with a short list.
- Filing-only monitoring. Dated, public and defensible, with a low false-positive rate on the observation itself. It also produces few events, so pipeline coverage will be thin unless the target universe is large and public.
- Multi-signal scoring with an evidence layer. Broadest coverage. The precondition is that each contributing signal keeps its own date, source and label rather than collapsing into one number before anyone reviews it.
None of these is correct in the abstract. The choosing question is how much research time the team can actually absorb, because a signal nobody verifies is a signal nobody used.
Four questions before buying another lead source
- Can you show me the source and the date for this specific lead? Not the methodology, the record. This question separates evidence from assertion faster than any other.
- Which of the three signal families produced this score, and in what proportion? If the answer is proprietary and unavailable, the buyer is purchasing a ranking rather than a reason.
- What happens when the underlying signal disappears? A withdrawn posting or an amended filing should update the record. Ask what the record looks like afterward.
- What is explicitly not claimed? The most useful line in a data contract is the one that says what the vendor does not guarantee.
CISA's Secure by Design guidance puts the surrounding obligation in producer-and-customer terms, stating that every technology provider must take ownership at the executive level to ensure their products are secure by design, and that the burden has been placed disproportionately on consumers and small organizations rather than on producers. It also publishes a Secure by Demand guide described as concise and usable by any customer of software during procurement discussions with third party resellers or service providers. That guide is written about software procurement rather than lead data, and the transfer to a data contract is this publication's reading rather than CISA's claim.
A 30-day check
- Days 1 to 5. Pull twenty leads your team acted on last quarter. For each, try to name the event, the date and the source without opening the vendor's interface.
- Days 6 to 10. Count how many of the twenty produced a verifiable source. That ratio is your current evidence rate, and it is the number worth improving.
- Days 11 to 15. Add publication date and source fields to the pipeline if they do not exist. Do not backfill them with guesses.
- Days 16 to 22. Rework one segment of the target list into the five-field format and compare representative reaction with the scored list.
- Days 23 to 30. Write the disqualification rule: what evidence would cause the team to drop an account, and who is authorized to record it.
What remains unconfirmed
- Any measured effect of evidence-backed qualification on reply rates, meetings or win rates. No benchmark or customer result was reviewed, and none is asserted.
- How DealFORGE weights its three signal families, and whether individual signal dates are exposed in the interface. No product walkthrough was conducted for this edition.
- Whether current commercial lead-data contracts typically include signal provenance or date fields. No contract was reviewed.
- Current JOLTS figures for any industry or month. The survey's structure was described; no data point was cited.
- Any specific company's hiring or filing activity. No company was examined for this edition, and the examples used are hypothetical.
- Market data. No stock, index or valuation figures were reviewed.
Evidence notes and limits
DealFORGE's approved product description is the source for the three signal families named in this article — buying signals derived from official filings, hiring activity and predictive business needs — and for the positioning of those results as evidence-based opportunities to research and contact rather than guaranteed purchases or closed accounts. The trial description is likewise the approved statement: at least 50 leads that participants can keep and work independently, with no obligation to purchase. No trial duration, credit-card requirement, signup path, refund term, lead-quality figure or customer outcome is asserted, because none is approved.
SEC EDGAR is cited for the general proposition that public company disclosure documents are filed through the system and carry a filing date and an accession number that make each submission individually retrievable. It is not cited for any specific company's filing, and no filing content is characterized here.
The Bureau of Labor Statistics' Job Openings and Labor Turnover Survey is cited for the proposition that the survey reports job openings, hires and separations as distinct measures rather than a single hiring figure. No JOLTS data point, seasonally adjusted series value or industry breakdown is quoted, and the application of that distinction to lead qualification is this publication's analysis rather than a BLS finding.
NIST's AI Risk Management Framework is cited solely for its four named Core functions: Govern, Map, Measure and Manage. No control identifier, maturity score or measured outcome is asserted, and the observation about unattributed scoring outputs is labeled as this publication's analysis.
CISA's Secure by Design page is the source for the quoted statement that every technology provider must take ownership at the executive level to ensure their products are secure by design; for the quoted statement that the burden has been placed disproportionately on consumers and small organizations rather than on producers; and for the existence and described use of the Secure by Demand guide in procurement discussions with third party resellers or service providers. It is voluntary guidance, and no certification or measured outcome is asserted from it. Applying a software-procurement guide to a lead-data contract is this publication's reading.
No lead record, vendor contract, pipeline export, product demonstration or customer result was reviewed. The three signal families' differing clocks, the five-field record, the three assembly options, the four questions and the 30-day check are this publication's analysis. Nothing here establishes that any organization improved reply rates, reduced research time or closed more business.
The advertisement above is separate from this analysis and is not a source for any claim in it.