Intelligence Community Directive 203, Analytic Standards, establishes the tradecraft standards that all IC analytical products must meet. ICD 203 is the institutional embodiment of the analytical norms Sherman Kent established informally — codified into enforceable standards after the Iraq WMD failure demonstrated the cost of inadequate tradecraft discipline.

The nine analytic tradecraft standards

ICD 203 mandates that all analytical products must:

1. Be objective. Analysis must be free from politicization — political, personal, or institutional bias must not influence analytical judgments. Analysts must “call it as they see it,” regardless of policy preferences or consumer expectations.

2. Be independent of political considerations. Analysis must not be shaped by the policy positions of the executive or legislative branches. The analyst’s obligation is to the evidence, not to the policy the consumer prefers.

3. Be timely. Products must be delivered when the consumer needs them. An accurate assessment delivered after the decision point has no operational value.

4. Be based on all available sources of intelligence. All-source analysis — products must consider information from all collection disciplines, not rely on a single source. The Iraq WMD failure (heavy reliance on CURVEBALL’s unvalidated HUMINT) demonstrated the cost of single-source dependence.

5. Properly describe quality and reliability of sources. Each significant source must be characterized for reliability and access — can the source know what they claim to know? Have they been reliable in the past? Are there reasons to question the information? Source reliability evaluation prevents the kind of uncritical source reliance that produced the CURVEBALL problem.

6. Properly caveat and express uncertainties. Analytical judgments must be accompanied by explicit statements of confidence and uncertainty using estimative language standards. “We assess with high confidence” means something specific (see the estimative language standards entry); the analyst must use the correct language and the consumer must understand what it means.

7. Properly distinguish between underlying intelligence and analysts’ assumptions and judgments. Products must clearly separate what the evidence shows from what the analyst infers. Assumptions must be identified as such — the key assumptions check is the structured analytic technique designed to surface hidden assumptions.

8. Incorporate analysis of alternatives. Products must consider alternative explanations and outcomes, not simply build a case for the most likely explanation. This standard — the institutional response to the groupthink problem — requires that products address “what if we’re wrong?” Analysis of competing hypotheses is the formal method for meeting this standard.

9. Demonstrate relevance to U.S. national security. Products must address questions that matter — the standard prevents analytical resources from being consumed by interesting but operationally irrelevant topics.

Confidence levels

ICD 203 standardizes the confidence language used in analytical products:

  • High confidence — judgments based on high-quality information and/or the nature of the issue makes it possible to render a solid judgment. Does not mean certainty.
  • Moderate confidence — information is credibly sourced and plausible but not sufficient to warrant high confidence. Or the issue is inherently uncertain.
  • Low confidence — information is fragmentary or poorly corroborated, or the issue is inherently highly uncertain. Low-confidence judgments are still intelligence community judgments — they represent the best assessment available given the evidence.

Post-Iraq reform context

ICD 203 was issued in 2007, replacing earlier standards, in direct response to the Iraq WMD analytical failures. The Silberman-Robb Commission (2005) and the Senate Select Committee on Intelligence inquiry documented specific tradecraft failures — insufficient alternative analysis, uncritical source reliance, assumption persistence — that the nine standards are designed to prevent.

The question — which this vault’s analysis of intelligence failure addresses — is whether tradecraft standards can prevent structural failures or only individual ones. ICD 203 can ensure that analysts consider alternatives, evaluate sources, and caveat judgments. It cannot ensure that the system’s categories encode the adversary’s relevant properties — the legibility limitation that tradecraft standards do not address.