A list is inventory with a date on it. Most organisations record the date they bought the record, not the date anyone last confirmed it.

That distinction is invisible in a weekly report and decisive in a quarterly review. Bounce rates catch the records that are loudly wrong. Nothing catches the record that still resolves and no longer describes reality.

What this article will not use

Decay benchmarks are widely published and almost entirely vendor-published. The figures in circulation disagree with one another, and none could be verified against the sources this publication treats as checkable.

So no decay percentage appears below. The argument does not need one, because the more useful evidence is a public record whose own metadata can be inspected line by line.

A public record and its timestamps

The SEC's EDGAR full-text search covers the full text of filings submitted electronically since 2001, including attachments such as exhibits. Its documentation is unusually specific about what the tool does and does not do.

It searches by keyword, ticker, company name, CIK number or reporter's last name, individually or in combination. It supports implied AND, exclusion, OR, proximity with NEAR, exact phrases in quotation marks, and trailing wildcards.

The same help page states that natural language search capabilities are not currently supported. That sentence is worth keeping. A researcher who treats a search result as an answer is using a Boolean engine as though it were a conversation.

Filters narrow by form type and by date range, including a custom start and end date. Results are ranked by relevance to the query — a helpfulness measure, not a quality judgement about the filing.

The same publisher, three different coverage statements

Here is the detail that makes the case better than any vendor statistic.

The full-text search FAQ says coverage runs from 2001. The SEC's data resources page, marked as last reviewed or updated in September 2026, describes the same tool as searching the full text of EDGAR filings “from the last four years.” An older EDGAR search updates page, dated 2006, describes a beta version covering “the last two years.”

None of that is scandalous. Documentation ages in place, and descriptive text routinely outlives the thing it described. But it produces a concrete rule for anyone building an evidence file: cite the page, not the domain.

“The SEC says” is not a citation. A specific page, with its own review date, is.

Why the bounce is the cheap failure

A hard bounce costs one send and some sender reputation. The damage is real, it is measurable, and it arrives immediately.

What makes it cheap is not that it is harmless. It is that it is loud, and loud problems get fixed. Delivery tools flag it, lists get scrubbed, and the team knows within a day.

Relevance decay has no equivalent instrument. Nothing bounces when a contact changes title. Nothing bounces when the buying committee is restructured out of the decision.

The expensive failure is relevance

Picture the work: a representative researches an account, reads the most recent filing, drafts a considered first approach, and sends it to a person whose remit changed last quarter.

The message delivers. Nothing bounces. The effort was simply aimed at a version of the account that no longer exists — and unlike a bounce, nobody will report it.

This is why list quality and lead qualification are the same problem wearing two job titles. Verification asks whether you can reach someone. Qualification asks whether there is a current reason to.

What a defensible record must carry

If the goal is a record you can defend in a review, four fields do most of the work.

  • The claim. One sentence, specific enough to be wrong.
  • The source. An exact page, not a domain and not a vendor's verified badge.
  • The source's own date. When the underlying evidence was published, which is frequently not when you found it.
  • The check date and the owner. When a named person last confirmed it, and who carries that accountability.

Three of those four are metadata, and metadata is exactly what a purchased export tends to strip out.

Where DealForge fits, in its own approved terms

JWilliams & Associates describes DealForge as finding current B2B lead opportunities using buying signals from official filings, hiring activity and predictive business needs.

Those are positioned as evidence-based opportunities to research and contact, not guaranteed purchases or closed accounts.

Read that against the four fields above and the standard being set is the first two: a signal, and a source someone can open. Official filings satisfy that test more easily than most because the underlying document is public and carries its own filing date.

Hiring activity and stated business needs are likewise signals to verify rather than facts to assume. A job posting is evidence that a function is being built, not proof of what it will buy.

No lead-quality figure, conversion rate or customer outcome is claimed here. Any such number would require a measured baseline in a specific market.

Six decision criteria

  • Verified-at date. Does every record carry the date it was last checked, or only the date it was acquired?
  • Field-level freshness. Is the contact channel treated as decaying faster than the firmographic, or is the whole record given one age?
  • Signal provenance. When a record is flagged as an opportunity, can you see what evidence triggered it and when that evidence was published?
  • Degradation handling. When a contact turns invalid, is it deleted, or is it treated as a possible job change worth re-routing?
  • Verification placement. Does verification run before each send, on a calendar, or only when bounce rates spike?
  • Owner. Is there a named person accountable for reachable share, or is data quality everyone's job in principle and nobody's in practice?

Three options and their real costs

Buy a refreshed list each quarter. Fast, familiar, and it buys discovery in segments you do not already cover. The cost is that the export starts ageing the moment it lands, arriving with a vendor's historical badge attached to records nobody has probed since.

Maintain your own base. Verification on entry, a stored verified-at date, and a re-check cadence that touches every record within a defined window. The cost is cadence discipline, which is precisely what a busy quarter erodes first.

Monitor buying signals instead of lists. Rather than maintaining a population, watch for evidence that a specific account has a current reason to talk. The cost is narrower coverage and a higher bar per record: signals are sparser than names, and each one still needs a reachable person behind it.

Most teams need two of the three. Discovery without maintenance produces a base that decays in place. Maintenance without discovery produces a shrinking asset.

A 30-day pilot that scores precision, not volume

  • Day 1 to 5, measure what you have. Take 200 to 500 records at least 12 months old, verify them, and calculate the share that changed. That is your own decay rate, and it beats any published benchmark.
  • Day 6 to 15, define 'current'. Write down what counts as a current reason to contact an account. If the answer is 'they are in our target industry', you have a list, not a qualification.
  • Day 16 to 25, run both and compare. Pull one batch from a refreshed list and one from a signal-driven approach. Score each on whether you can name the evidence, its date and a reachable person.
  • Day 26 to 30, publish the result. Report precision and provenance, not lead count. A hundred records you can defend beat a thousand you cannot.

Four indicators worth baselining

Reachable share: confirmed-valid records divided by all records. Provenance coverage: the share of flagged opportunities with a dated, checkable source. Verification currency: the share of active records confirmed within your defined window. Signal-to-meeting rate: meetings held per hundred signals acted on.

Baseline all four on one segment and one quarter. An improvement claim built from your own before-and-after is defensible. A vendor's decay percentage is a planning assumption, not your number.

Evidence notes and limits

Evidence was reviewed through 20:00 UTC on November 1, 2026.

Four sec.gov pages support the factual claims above: the EDGAR full-text search FAQ, the EDGAR advanced search interface, the SEC data resources page marked as reviewed or updated in September 2026, and the EDGAR search updates page dated 2006. All are the regulator's own documentation.

Those pages describe the same search coverage differently, and the article reports the discrepancy rather than resolving it. The FAQ's 2001 statement is the most specific; the data resources page's four-year phrasing appears alongside newer material and is treated as older descriptive text.

Earlier versions of this article relied on vendor-published decay research that could not be verified against approved sources. Those figures are removed rather than restated, and no replacement statistic was invented.

DealForge claims are limited to the owner-approved description and are not characterised as verified outcomes. No customer results, savings, conversion rates or comparative performance claims are asserted. The criteria, options, pilot and indicators are analysis and suggested method, not a tested protocol.