Structured Analytic Techniques (SATs)

BLUF. Structured Analytic Techniques are procedures that force analysts to externalise reasoning steps that, in unstructured analysis, remain implicit and therefore unauditable. SATs do three things simultaneously: they externalise judgment so it becomes visible to peer review; they debias by interrupting heuristic shortcuts that produce systematic error; and they audit by leaving a procedural record that allows post-hoc reconstruction of how a judgment was reached. SATs are not a replacement for analyst expertise — they are scaffolding around expertise that makes it accountable. This guide catalogues the doctrinal SAT taxonomy, deep-dives the three most consequential techniques (ACH, Key Assumptions Check, Red Team), and maps SATs into the modern OSINT and LLM-assisted analytic pipeline.

Doctrinal Foundation

The canonical reference is Richards J. Heuer Jr. and Randolph H. Pherson’s Structured Analytic Techniques for Intelligence Analysis (3rd ed., CQ Press, 2021), which codifies the techniques originally distributed across CIA training materials and Sherman Kent School publications.

Doctrinal mandate. ODNI’s Intelligence Community Directive (ICD) 203 — Analytic Standards (2015) requires that IC analytic products be “objective, independent of political consideration, timely, based on all available sources of intelligence, and exhibit proper standards of analytic tradecraft.” ICD 203 does not name specific SATs but mandates the analytic outcomes SATs are designed to produce. The UK Professional Head of Intelligence Assessment (PHIA) maintains an equivalent doctrinal framework, with broadly comparable technique catalogues.

Why SATs exist. The motivating empirical claim, established by Heuer’s Psychology of Intelligence Analysis (1999) and Kahneman-Tversky’s heuristics-and-biases program, is that unstructured expert judgment is systematically biased in identifiable ways (see Cognitive Biases in Intelligence Analysis). SATs are procedural countermeasures whose efficacy derives precisely from their structural friction — they slow the analyst down at the moments where bias operates fastest.

SAT Taxonomy

The standard Heuer-Pherson taxonomy distributes techniques across three categories by analytic function.

Diagnostic Techniques — Surface what is already in the evidence

Technique	Purpose	When to Use
Key Assumptions Check	Surface implicit assumptions underlying the prevailing assessment	At the outset of any major assessment; after surprise
Chronologies / Timelines	Establish temporal order, identify causation gaps, surface anomalies	Any event-driven analysis; investigations with multiple actors
Indicators Validation	Pre-define observable signals that would confirm or refute a hypothesis	I&W work; long-running investigations with multiple branches
Starbursting	Generate the full who/what/when/where/why/how question set	Early-stage problem framing; intelligence gap identification
Network / Link Analysis	Map relationships between entities, surface bridge nodes and clusters	Actor/organisation analysis; financial flows; influence operations
Mind Mapping	Externalise the analyst’s mental model of a problem space	Onboarding to a new account; complex topic decomposition

Contrarian Techniques — Force engagement with the alternative

Technique	Purpose	When to Use
Analysis of Competing Hypotheses (ACH)	Systematically grade all evidence against all hypotheses, surfacing the least-disconfirmed	Mid-stage analysis once hypotheses and evidence are inventoried
Devil’s Advocacy	One analyst argues against the consensus regardless of personal view	When consensus crystallises early or groupthink risk is high
Red Team	Adversary-perspective team challenges friendly assumptions	War-gaming, exercise design, counterintelligence, deception detection
Pre-Mortem	Imagine the assessment has failed; reason backward to identify failure path	Before final dissemination of high-stakes assessment
High-Impact / Low-Probability (HILP)	Force attention on scenarios discounted as improbable but high-consequence	Strategic warning, black-swan preparation
What-If Analysis	Bypass probability estimation; ask what indicators would precede a given outcome	Adversary capability assessment; warning indicator design

Imaginative / Synthesis Techniques — Generate hypotheses not currently in the frame

Technique	Purpose	When to Use
Brainstorming (structured)	Generate divergent hypotheses before convergent evaluation	Problem framing; new accounts; unfamiliar adversaries
Outside-In Thinking	Force analysis of external/environmental drivers acting on the target	Strategic-level assessments; long-horizon forecasts
Morphological Analysis	Decompose problem into orthogonal dimensions; enumerate combinations	Capability assessments; campaign-design scenarios
Alternative Futures Analysis	Construct 2–4 internally consistent future scenarios	Strategic estimates; long-horizon planning
Cross-Impact Matrix	Map how the occurrence of one event changes probabilities of others	Multi-variable strategic assessments; scenario coupling
Delphi Method	Iterated anonymous expert elicitation toward calibrated consensus	Specialist-judgment aggregation; deep technical assessments

Deep Dive: Analysis of Competing Hypotheses (ACH)

ACH is the single most consequential SAT in the modern catalogue. The procedure:

1. Enumerate all plausible hypotheses (≥3, typically 4–7). Include hypotheses the analyst considers unlikely.
2. Inventory the available evidence and assumptions, item by item.
3. Construct a matrix: hypotheses across the top, evidence down the side.
4. For each cell, mark whether the evidence is consistent (C), inconsistent (I), or not applicable (N/A) to that hypothesis.
5. Count inconsistencies per hypothesis, not consistencies. The least-disconfirmed hypothesis is the one with the fewest “I” marks.
6. Sensitivity-test: which evidence items, if reversed in reliability, would change the ranking? Those items drive collection priorities.

Why ACH works: unstructured analysis evaluates hypotheses serially — analyst proposes a hypothesis, looks for confirming evidence, and stops. ACH evaluates evidence against all hypotheses simultaneously, making suppressed disconfirming evidence visible. It also reframes the analytic question from “what is the most likely hypothesis” to “which hypothesis is least disconfirmed” — a small linguistic shift that interrupts confirmation bias.

Limitations. ACH cannot manufacture hypotheses the team has not considered — its output is bounded by the hypothesis set at step 1. It is also evidence-quality-blind: low-reliability evidence weighted equally with high-reliability evidence produces noise. Mitigation: gate ACH on a prior source-reliability audit (see Source Verification Framework).

See ACH for the dedicated procedural reference.

Deep Dive: Key Assumptions Check

The Key Assumptions Check is the lowest-cost, highest-yield SAT in the catalogue. It is also the most frequently skipped.

Procedure (10–15 assumption step):

1. The team brainstorms 10–15 assumptions underlying the current assessment. Assumptions are propositions taken to be true without being separately argued.
2. Each assumption is written in declarative form (“we are assuming that X”).
3. For each assumption, the team answers four questions:
   • Why are we confident in this assumption?
   • What evidence supports it?
   • What would falsify it?
   • If it were false, how would the assessment change?
4. Assumptions are graded: Solid, Solid with caveats, Unsupported, or Falsified.
5. Unsupported or falsified assumptions trigger explicit rework of the assessment.

Why it matters. Most major intelligence failures, on retrospective inspection, trace to one or two assumptions never identified as assumptions. The pre-1973 Israeli “Concept” was a cluster of assumptions about Egyptian preconditions for war that were never separately tested (see Cognitive Biases § Confirmation Bias and Yom Kippur War).

Deep Dive: Red Team vs. Devil’s Advocacy

The two are routinely conflated and are doctrinally distinct.

Devil’s Advocacy is an internal technique: one analyst (or sub-team) within the producing organisation is assigned to argue against the consensus regardless of personal view. Its purpose is procedural — to ensure the consensus has been challenged on the record before dissemination.

Red Team is an adversary-perspective technique: the team adopts the adversary’s worldview, decision calculus, and information environment, and reasons forward to the adversary’s most probable actions. Its purpose is substantive — to test whether friendly assumptions about adversary behaviour survive contact with a serious effort to think as the adversary.

The distinction matters because they produce different outputs. Devil’s Advocacy produces the strongest counterargument to the friendly consensus. Red Team produces the adversary’s most probable course of action — which may or may not contradict the friendly consensus. Both are necessary; neither substitutes for the other.

Operational note. Red Teaming is methodologically demanding: it requires genuine cultural-linguistic fluency in the adversary’s frame. A Red Team staffed with analysts who only know the adversary through Western secondary sources will mirror-image inside the exercise and produce a Red Team report that recapitulates the friendly consensus in adversary costume. See Cognitive Biases § Mirror-Imaging.

SATs in the OSINT Pipeline

Each phase of the Intelligence Cycle applied to OSINT has a default SAT mapping:

Phase	Default SAT	Function
Direction / Planning	Starbursting, Key Assumptions Check	Frame PIRs; surface assumptions in tasking
Collection	Indicators Validation	Pre-define what evidence would confirm/refute each hypothesis
Processing	Source Verification (not formally a SAT but doctrinally adjacent)	Grade source reliability before evidence enters analysis
Analysis	ACH, Network/Link Analysis, Chronologies	Hypothesis testing, relational mapping, temporal reconstruction
Production	Red Team, Devil’s Advocacy, Pre-Mortem	Stress-test draft before dissemination
Dissemination / Feedback	Lessons-learned chronology	Build the post-hoc audit trail

SATs and LLM-Assisted Analysis

LLMs can partially automate several SATs: enumerating hypotheses, generating Key Assumption candidates, building chronologies from raw text, drafting Red Team perspectives. This is operationally valuable.

The structural risk: LLM-generated alternative hypotheses are sampled from the model’s training distribution, which reflects the dominant analytic consensus on the topic. The LLM is therefore more likely to surface mainstream alternatives than the genuinely contrarian ones a human Red Team would identify. Automating the bias-mitigation step risks automating the bias — particularly mirror-imaging, since LLMs are trained predominantly on English-language sources that themselves mirror-image.

Mitigation. LLMs as SAT scaffolding require:

• Native-language sourcing as input, not English secondary sources
• Explicit “adversary-frame” prompting that names the adversary’s worldview before requesting analysis
• Human review of LLM output against an independent Red Team brief
• Confidence-language calibration imposed after LLM generation, not adopted from it

See LLM-Assisted OSINT SOP (A2IC) for the operational protocol.

Limitations of the SAT Programme

• The “structured theatre” critique. Stephen Marrin and others have argued that SATs are ritually performed without changing the underlying judgment, producing the appearance of structured analysis as cover for unstructured intuition. Empirical evaluations of SAT efficacy are sparse.
• Time cost. Full SAT application to a single assessment is days of senior-analyst time. Production tempo in current-intelligence settings rarely permits it.
• Hypothesis-set selection bias. SATs operate over whatever hypotheses make it into step 1; they cannot surface what no one has considered.
• Evidence-quality ceiling. SATs cannot upgrade weak evidence; they only structure judgment over the evidence in hand. The Iraq WMD NIE underwent SAT-equivalent procedures and still failed, because the underlying HUMINT (notably “Curveball”) was fabricated and the verification discipline upstream of SAT application was inadequate.

Key Connections

• ACH — dedicated procedural reference
• Cognitive Biases in Intelligence Analysis — failure modes SATs counter
• Intelligence Confidence Levels — calibrated language layer over SAT outputs
• Source Verification Framework — evidence-quality discipline upstream of SATs
• Intelligence Cycle — workflow context for SAT placement
• Intelligence — parent concept
• Intelligence Failure — failure-mode lens
• Indications and Warning — discipline heavily reliant on Indicators Validation
• Social Media Intelligence, Cyber Threat Intelligence, Financial Intelligence — OSINT disciplines where SATs apply
• OSINT — open-source context
• LLM-Assisted OSINT SOP (A2IC) — LLM/SAT integration
• Attribution — domain where ACH and Red Team are routine
• Richards J. Heuer Jr. — doctrinal author

Sources

• Richards J. Heuer Jr. and Randolph H. Pherson, Structured Analytic Techniques for Intelligence Analysis, 3rd ed. (CQ Press, 2021) — High confidence
• Richards J. Heuer Jr., Psychology of Intelligence Analysis (CIA CSI, 1999) — High confidence
• ODNI, Intelligence Community Directive (ICD) 203 — Analytic Standards (2015) — High confidence
• UK Professional Head of Intelligence Assessment (PHIA), Professional Development Framework — Medium confidence (limited public documentation)
• David T. Moore, Critical Thinking and Intelligence Analysis (National Defense Intelligence College, 2006) — High confidence