These standards apply to all analytical work produced within the School, in coursework, assessment, supervised research, and any material prepared for institutional clients. Their purpose is narrow and demanding: judgments must be traceable, honest about uncertainty, and defensible to a reader who disagrees with them.
Standards of this kind exist because analytical failure is rarely caused by missing information. It is usually caused by reasoning that was never made visible, so nobody could inspect it. A judgment whose evidence, assumptions, and confidence are all stated can be challenged, corrected, and improved. A judgment presented as a conclusion cannot.
Core requirements
Nine requirements apply to every analytical product.
Distinguish fact, inference, judgment, and assumption. These are four different things and must be separable by the reader. A fact is observed and sourced. An inference follows from evidence. A judgment adds analytical weighting. An assumption is unverified and load-bearing. Collapsing them into one confident voice is the most common failure in analytical writing.
Assess source reliability and information credibility separately. How dependable is this source in general, and how believable is this particular claim? The two answers diverge constantly. A generally reliable source can be wrong on a specific point, and an unreliable source can report something true.
Seek corroboration from independent sources. Independence is the operative word. Three outlets reproducing one wire report constitute one source, not three.
State confidence levels and probability language explicitly. Vague estimative language transfers the analyst’s uncertainty to the reader without disclosing it.
Consider alternative hypotheses. At least one serious alternative explanation must be examined, not a token straw man built to lose.
Identify cognitive bias and key assumptions. Name them in the product. Assumptions that go unnamed are the ones that fail silently.
Acknowledge uncertainty and record dissenting analysis. Where reviewers disagreed, the disagreement is documented rather than smoothed away. Suppressed dissent is a recurring feature of documented intelligence failures.
Maintain an audit trail and require human approval. Every claim must be traceable to its source, and no analytical product is released without human review. This applies without exception to AI-assisted work.
Revise and correct when new evidence emerges. A superseded judgment is corrected openly, with the change recorded. Analysis is a standing position, not a publication.
Provisional confidence scale
The School uses a three-level confidence scale, stated explicitly in every assessment.
High confidence. Supported by multiple reliable and independently corroborated sources, with limited major uncertainty remaining.
Moderate confidence. Supported by credible sources with partial corroboration, where meaningful gaps or uncertainties remain.
Low confidence. Based on limited, conflicting, indirect, or weakly corroborated evidence.
Low confidence is a legitimate finding, not a failure. An analyst who reports low confidence accurately has done the job. An analyst who inflates confidence to appear useful has not.
Confidence is not probability
This distinction is routinely mishandled and is worth stating on its own.
Confidence expresses how much trust we place in the evidence and reasoning. Probability expresses how likely an event is to occur. They are independent dimensions. You can hold high confidence that an outcome is unlikely, when the evidence is strong and points clearly away from it. You can hold low confidence in a probable outcome, when the evidence is thin but leans one way.
Stating a probability without a confidence level tells the reader what you think but not how much weight to put on it. Stating confidence without probability tells the reader you are certain about something unspecified. Analytical products from this School state both.
Report structure
All analytical reports produced in the School follow a standardized structure. The sequence is fixed so that a reader can find any element without hunting, and so that no required element can be quietly omitted.
Every report includes an executive summary; the situation; the evidence; the sources; the confidence assessment; at least one alternative explanation; the indicators being monitored; the forecast; the implications; the recommendations; the limitations; and a bibliography.
Two of those deserve emphasis because they are the ones most often dropped under time pressure.
Alternative explanation is mandatory. A report presenting a single explanation has not demonstrated that the explanation was tested.
Limitations is mandatory. Every assessment has boundaries in coverage, source access, timeframe, and method. Stating them is what allows a decision-maker to judge how far the analysis can be pushed.
The executive summary carries the judgment and its confidence level in the opening lines. A summary that describes the topic instead of stating the finding has failed at its only job.
Application in teaching and assessment
These standards are taught progressively rather than announced. Foundation courses introduce the fact-inference-judgment-assumption distinction and source-versus-information assessment. Method courses at Level 2 build structured technique, estimative language, and alternative hypothesis testing. OSINT courses at Level 3 add documentation and audit-trail requirements. Applied courses at Levels 4 and 5 require the full report structure. The progression maps to the School’s academic clusters and is reflected in the course catalog.
Assessment applies the standards directly. Applied work is marked on whether the reasoning is traceable, whether confidence is justified by the evidence presented, whether a genuine alternative was tested, and whether limitations are stated honestly. A well-written assessment with inflated confidence scores lower than a cautious one with defensible confidence.
Relationship to ethics and governance
Analytical standards and ethical boundaries operate together. Standards govern how reasoning is conducted; ethics governs what may be collected and how findings may be used. Rigorous method applied to unlawfully obtained material is not acceptable work, and lawfully collected material analysed carelessly is not acceptable either. The collection and conduct rules are published under ethics and legal boundaries.
The requirement of human approval sits in both frames. Methodologically, it exists because automated output is not traceable reasoning. Ethically, it exists because accountability for a judgment must rest with a person. Neither justification weakens as tools improve.
Analytical products published externally by the KBA13 Intelligence Center follow the same standards, which is what makes sanitized Center material usable as teaching cases. Learners see the standard applied to real problems rather than described in the abstract. Further context is available on the about the School page, the programs page, and the FAQ.