Does an agent’s failure to find a fact mean the fact is false?

Use four different evidence states

In an invented research task, an agent is looking for a company’s customer attrition measure. It might encounter an explicit zero, a statement that the measure was not tracked, no result in its chosen documents, or a document it could not retrieve. Those states answer different questions.

Do not compress them into one empty cell. Use separate labels: explicitly reported value; explicitly unmeasured or unavailable measure; not found within the search; and access or retrieval incomplete. Attach the relevant period and scope to every label.

Build an original evidence ledger

Use the following hypothetical rows. None describes an actual company, search, or product result.

• Document A states zero qualifying incidents for a defined quarter: record the reported zero with that definition and period.
• Document B states the measure was not tracked: record not measured; do not substitute zero.
• A search of Documents C and D returns no matching phrase: record the documents and terms searched; the underlying value remains unknown.
• Document E cannot be retrieved: record the retrieval gap; do not describe it as a negative finding.

Keep an explicit zero specific

A reported zero needs the same definition checks as any other number. Zero qualifying events in one subsidiary is not a group-wide statement about every event. A rounded table value also needs its scale and any disclosed rounding convention before the reader treats it as exact absence.

SEC staff’s interactive-data FAQ discusses narrative no or none as a numerical amount when substituting zero preserves the meaning. Its XBRL guide separately distinguishes nil from zero and from not reporting a fact. These technical distinctions reinforce the need to inspect the actual disclosure, rather than filling every missing field with a numeric default.

Make the search claim reproducible

For the unsuccessful hypothetical search, save the question, document set, date range, terms used, sections inspected, and access failures. State the narrow result: no matching disclosure was located in these materials using this method. Avoid claiming an exhaustive investigation unless the coverage actually supports that claim.

Then name the next useful evidence request. It might be a definition in a note, another reporting period, or confirmation that the metric is not disclosed. Searching more pages is useful only if the added coverage addresses the gap; repeated empty searches do not become a reported zero.

Carry the unknown into the decision record

If a rule requires that measure, the evidence ledger should let the reviewer see that the rule cannot yet be evaluated. Whether to pause, request evidence, or use an accepted fallback is a separate policy choice. Do not call the company a failure merely because the agent lacked the necessary document.

Conversely, an unresolved gap is not positive support for a preferred thesis. Preserve uncertainty in both directions. An explicit negative finding, when one is actually supported, should still identify the tested claim and the evidence that contradicts it.

Method, sources, and limits

The ledger and cases are original hypothetical examples. SEC staff sources were opened and checked on 26 October 2026; they are technical staff guidance, not a determination about a real disclosure. No live issuer search or agent test was conducted for this article, and an unknown value is not an investment recommendation.