A standardised vocabulary for expressing likelihood and confidence in analytical products, so judgements are calibrated and consistently understood (Sherman Kent tradition, adapted for cyber).
Likelihood / probability scale (future or uncertain events)
| Phrase | Range |
|---|---|
| Almost certain | ≈ 95–99% |
| Very likely | ≈ 80–95% |
| Likely | ≈ 60–80% |
| Roughly even chance | ≈ 40–60% |
| Unlikely | ≈ 20–40% |
| Highly unlikely / Remote | ≈ 1–20% |
Confidence in the judgement itself
- High confidence — multiple high-quality, independent sources; strong corroboration; limited alternative explanations.
- Moderate confidence — credible sources but some gaps, single-source dependency, or plausible alternatives.
- Low confidence — limited, fragmented, or low-credibility information; significant uncertainty remains.
Usage rules
- State confidence explicitly with every major judgment.
- Example: "We assess with moderate confidence that Actor X is preparing a new phishing campaign targeting the financial sector."
- Avoid absolute statements when evidence is incomplete; avoid vague intensifiers ("very sophisticated") without support.
- Prefer "we assess / we judge / evidence indicates" over "we believe".
This vocabulary pairs with the wiki's confidence frontmatter field — the page-level analogue of per-judgment confidence. See also Admiralty Rating for source-level grading.
Related
- Bottom Line Up Front — confidence joins the BLUF structure
- Admiralty Rating — source-level grading of individual reports
- Structured Analytic Techniques — generate the well-tested judgments these scales express