Evaluate the source. Calibrate the claim.

Source reliability concerns the origin: its access, methods, expertise, track record, and possible limitations. A reputable publisher can still repeat an error, and a generally dependable source may be poorly placed to know about a particular event.

Information credibility concerns the specific material and claim: whether the account is first-hand, coherent, in context, current, and supported by evidence. A source can be reliable in one domain but not another; a plausible claim is not established merely because its source is familiar.

Keep the two judgments visible in your notes. The Office of the Director of National Intelligence's public sourcing directive describes source-quality factors such as accuracy, completeness, currency, access, validation, motivation, bias, and expertise. Those factors help readers judge a source; they do not decide whether every individual claim from it is true.

Look for other evidence that bears on the same claim, then ask whether it is genuinely independent. Several articles may all rely on one press release, interview, database, or anonymous post. That is multiple publications, but not necessarily multiple lines of evidence.

Compare records with different origins or methods where appropriate. Check what agrees, what conflicts, and what each source could not observe. Seek disconfirming information and plausible alternative explanations, not only evidence that supports an early theory. Corroboration strengthens support; it does not eliminate the possibility of shared error or missing context.

Preserve a short evidence trail: the claim being tested, the sources actually consulted, the important agreements or disagreements, and any unresolved gap. The ODNI's Analytic Standards likewise call for clear reasoning, attention to contrary information and alternatives, and an explanation of uncertainties in major judgments. These are public U.S. Intelligence Community standards, not a claim that CN-SIGMA is part of or endorsed by the government.

A judgment about an outcome and confidence in the evidence for that judgment are different things. If you use a confidence label, explain what it describes and why it fits. Confidence should reflect the strength and consistency of the evidence, the quality of the sources and methods, material assumptions, and the remaining gaps—not the analyst's certainty of tone.

Use plain language tied to the record. For example: “We assess X with moderate confidence because two relevant records support it, but both depend on the same underlying report and the date remains unconfirmed.” If you use a likelihood scale or numeric probabilities, define the scale and use it consistently; do not imply measurement precision that the evidence cannot support.

Name what could change the assessment: a missing document, independent confirmation, a conflicting account, or a new observation. Revisit the judgment when material evidence changes. Confidence language is most useful when a reader can see its basis and its limits.