Military Intelligence Analyst

Purpose

Commanders decide on violence from an incomplete, deliberately deceptive picture. The analyst turns fragments — an intercepted call, a satellite frame, a defector's story, a pattern in open reporting — into the clearest honest assessment of what the enemy can do and is likely to do. The discipline exists because the enemy hides, lies, and adapts, and because the worst failure is being confidently, unanimously wrong, foreclosing the commander's options.

Core Mission

Reduce the commander's uncertainty about the adversary and environment to the lowest defensible level; where it remains, state it with calibrated confidence so a guess is never taken for fact.

Primary Responsibilities

The visible work is assessments and briefings; the real work is disciplined reasoning against an adversary reasoning back. An analyst defines what the commander most needs to know (priority intelligence requirements); tasks collection; fuses contradictory reporting into a coherent picture; tests competing hypotheses rather than a favorite; produces indicators and warnings before enemy action and battle damage assessment after; grades source and information separately; and communicates in estimative language. Underneath it all is separating the known from the assessed from the hoped.

Guiding Principles

• Tell the commander what they need to know, not what they want to hear. Credibility is the analyst's only product, and it dies when the assessment bends to the command climate.
• Distinguish evidence from inference from assumption — out loud. An assumption stated as fact is the most dangerous line in a briefing.
• The enemy is not you. Mirror-imaging — assuming the adversary shares your logic, values, and risk tolerance — is the original sin.
• Confidence is a claim you must earn and quantify. "Likely" and "almost certain" report how good the evidence is.
• Look for what would prove you wrong. Rigor hunts the disconfirming indicator; bias fits data to the story.
• Grade the source and the information separately. A reliable source can pass a rumor; an unreliable one can witness the truth.
• Absence of evidence is not evidence of absence — but it is information. What you can't see is part of the assessment.

Mental Models

• The intelligence cycle. Direction, collection, processing, analysis, production, dissemination, feedback — a loop, not a line.
• Analysis of Competing Hypotheses (ACH, Heuer). Score each item by how inconsistent it is with each hypothesis; you disconfirm rather than confirm, and the survivor is the one evidence cannot refute.
• Indicators and warnings (I&W). Pre-identify signatures that precede enemy action, so warning precedes the event.
• Intelligence Preparation of the Battlefield (IPB). Define the environment, its effects, the threat, and the threat's likely courses of action.
• The cone of plausibility. Project along a widening cone of futures; baseline up the middle, wildcards the edges — forcing attention to the outlier.
• Source reliability grading (Admiralty / NATO code). Rate source A–F and information 1–6 independently; "A1" and "F6" are distinct.

First Principles

• All collection is partial, and the gaps are not random — the enemy chose them.
• Deception is the default against a capable adversary, not the exception.
• A single source confirming itself is not corroboration.
• The map of the enemy is a model; all models are wrong, so ask whether this one is useful and honestly bounded.
• Being precisely wrong is worse than roughly right; precision invites trust.

Questions Experts Constantly Ask

• What does the commander need to decide, and does this answer that?
• What are all the hypotheses, including the uncomfortable ones?
• What evidence would I expect if my favored hypothesis were false?
• Is this corroboration, or one source laundered through three reports?
• Am I assuming the enemy will act as I would? Where am I mirror-imaging?
• How reliable is this source, and how credible is this report?
• What's the most dangerous course of action, even if not the most likely?
• What is my confidence, and can I defend that word if challenged?

Decision Frameworks

• Priority Intelligence Requirements (PIR). Translate the commander's decisions into the few questions whose answers change a decision; subordinate collection to them, and let nothing else win it.
• Most likely vs. most dangerous COA. Brief both; planning against the likely alone leaves the force exposed to the worst.
• Estimative language (Sherman Kent). Map words to probability bands — "remote," "unlikely," "even chance," "likely," "almost certain" — a number per word.
• Key assumptions check. List the load-bearing assumptions; for each, ask what breaks if it's false, and whether you'd notice.
• Multi-INT corroboration. Weight a judgment higher when independent disciplines (SIGINT, IMINT, HUMINT, OSINT) converge from separate collection; single-source judgments are fragile.

Workflow

1. Receive and refine the requirement. Pin down the real question; convert vague interest into PIRs tied to decisions.
2. Plan collection. Map each requirement to the discipline likeliest to answer it; request collection for the gaps.
3. Process the raw. Translate, geolocate, transcribe, and triage reporting; grade source and information separately.
4. Generate hypotheses. Lay out the full set of explanations or enemy COAs before reading evidence into one.
5. Test by disconfirmation (ACH). Score evidence by inconsistency; let the hypothesis the evidence cannot kill rise; check key assumptions.
6. Assess confidence. Assign estimative language reflecting source quality, corroboration, and gaps.
7. Produce and disseminate. Lead with the bottom line, then reasoning, confidence, and gaps, in a form the commander can act on.
8. Warn and update. Watch the indicators, reissue warnings as they trip, and revise as reporting lands.
9. Assess effect (BDA). Judge what was achieved against intent and feed it back into the cycle.

Common Tradeoffs

• Timeliness vs. accuracy. A perfect assessment after the decision is worthless, a fast wrong one dangerous. Deliver the best defensible answer by the deadline and flag what's uncertain.
• Breadth vs. depth of collection. Finite assets watch many things shallowly or few deeply; PIRs force the choice.
• Confidence vs. honesty. Commanders crave certainty; report real confidence when it's low.
• Protecting sources vs. sharing the picture. Revealing how you know can burn the source; withholding it leaves the consumer unable to weigh it.
• Most likely vs. most dangerous. Resourcing against the outlier costs effort; ignoring it invites surprise.

Rules of Thumb

• If every source agrees instantly, suspect a single origin or a deception.
• The report that perfectly confirms your theory deserves the most scrutiny.
• Name your assumptions; the unwritten ones bite.
• An honest "I don't know" beats a confident guess dressed as fact.
• Geolocation and timestamp before narrative — anchor reporting in space and time.
• A capability is not an intention; don't conflate "could" with "will."
• When the picture is too clean, you're either very good or being played — assume the latter until you rule it out.

Failure Modes

• Confirmation bias. Building the case for your first conclusion, discounting what contradicts it.
• Mirror-imaging. Assuming the adversary's rationality, values, and risk appetite match yours.
• Single-source dependence. Treating one well-placed source as truth until they're wrong, turned, or lying.
• Layering. Stacking assessment on assessment until the original evidence is buried and the chain can't be retraced.
• Politicization. Shading analysis toward the answer the command wants.
• Crying wolf or under-warning. So many alerts that warning is ignored, or such fear of error that none goes out.
• Vivid-evidence bias. Over-weighting the dramatic intercept over the duller pattern.

Anti-patterns

• The single-hypothesis brief — one explanation, as if no alternative existed.
• Estimative mush — "may possibly potentially" hedging that gives the commander nothing to act on.
• Source laundering — citing three reports tracing to one source as independent corroboration.
• Stovepiping — disciplines hoarding their take, so no one fuses the picture.
• Connecting dots that aren't there — imposing a narrative on noise because pattern feels like insight.

Vocabulary

• PIR — priority intelligence requirement; a question whose answer changes a decision.
• OSINT / HUMINT / SIGINT / IMINT — intelligence from open sources, human sources, signals, and imagery.
• All-source analysis — fusing all disciplines into one assessment.
• Indicators and warnings (I&W) — signatures that precede adversary action.
• BDA — battle damage assessment; effect judged against intent.
• Estimative language — probability words tied to fixed bands.
• Mirror-imaging — assuming the adversary thinks as you do.
• Admiralty code — the A–F / 1–6 grading of source reliability and information.
• EEFI — essential elements of friendly information; what we hide from the enemy.
• Order of battle — the composition, disposition, and strength of a force.

Tools

• All-source fusion platforms — to correlate reporting across disciplines.
• GEOINT/imagery exploitation tools — to measure, geolocate, and track overhead collection.
• Link and pattern analysis software — to map networks and activity.
• Structured analytic technique templates — ACH matrices, key assumptions checks, devil's advocacy, red teaming.
• The collection management matrix — mapping requirements to assets and gaps.
• The estimative language standard — keeping "likely" the same for writer and reader.

Collaboration

The analyst sits between collectors and commanders, trusted by neither automatically. They task HUMINT case officers, SIGINT and IMINT exploitation cells, and data scientists who find signal no human can read; brief planners and the commander without overselling the uncertainty; and share with partners under tight controls on sources and methods. The recurring friction is the operations–intelligence seam: operators want certainty now, analysts owe honesty about what they can't yet support. The best embed with the planners and aim at the decisions, not at unread reports.

Ethics

Intelligence informs decisions that kill, and an analyst who shades the truth — toward optimism, the command's preference, or a pet theory — can put violence on the wrong target. The first duty is intellectual honesty: state the known, assessed, and unknown without distortion, and resist politicization. Human sources can die if methods leak, so protecting them is a moral obligation. Collection and targeting touch non-combatants' lives, so the analyst owes diligence to distinction and to the cost of a wrong answer. When a consumer is about to act on a misread, speak up.

Scenarios

An imminent-attack warning that almost wasn't. SIGINT shows an armored brigade's logistics chatter spiking at a border; the cell reads it as a routine exercise. The analyst runs ACH: exercise, repositioning, and pre-attack staging all explain the chatter, but only staging explains the additional indicators — forward fuel, medical units moving up, emission control on the lead battalions. You'd broadcast a drill, not hide it. The analyst warns at "likely" with the disconfirming indicator named, rather than waiting for certainty that comes only with the attack — giving the commander a decision window the consensus would have shut.

A single dazzling source. A walk-in HUMINT source offers exquisite detail on enemy intentions, and operations staff want to act now. The analyst grades source and information separately: placement is plausible (B), but the information is uncorroborated (5) and suspiciously complete and flattering to the staff's preference — a deception signature. The assessment goes forward as "single-source, uncorroborated, possible deception," and collection is sent for independent confirmation first. The caveat is the product; without it you are being played.

Battle damage assessment under pressure. After a strike, the command wants to report the target destroyed. Imagery shows structural damage but no secondary explosions and no confirmed presence of the high-value individual. The analyst resists: "moderate confidence the facility is degraded; low confidence the target was present or killed; recommend re-strike authority held pending pattern-of-life confirmation" — the unwelcome answer that prevents a claim that collapses when the target reappears.

Related Occupations

The analyst reasons under uncertainty about an adversary. The data scientist shares the discipline of extracting signal from noisy data and quantifying confidence, but against indifferent nature, not a deceiving opponent. The cyber warfare specialist both consumes and generates intelligence on the same adversary, from inside their networks. The infantry officer is the analyst's most demanding consumer, fighting on the picture the analyst builds. The diplomat and policy analyst apply the same reasoning to statecraft, where the stakes are political, not tactical.

References

• Psychology of Intelligence Analysis — Richards J. Heuer Jr.
• Words of Estimative Probability — Sherman Kent
• Structured Analytic Techniques for Intelligence Analysis — Heuer & Pherson
• Intelligence Preparation of the Battlefield/Battlespace (ATP 2-01.3)
• Surprise Attack — Ephraim Kam