Why Intelligence Fails: Mind, Evidence, Organization

Why Intelligence Fails: Mind, Evidence, Organization

CommandEleven Intelligence Desk // Tradecraft Desk, Tradecraft Primer. Part of the series How CommandEleven Intelligence Thinks.

Bottom Line Up Front

Intelligence fails for reasons that are mostly ordinary. The previous primer showed where in the process failures were found. This one asks why they happen. The answers fall into three layers: the analyst’s mind, which filters and simplifies by nature; the evidence, which is incomplete, noisy and sometimes planted; and the organization, which decides what is asked, shared and challenged.

The central point comes from Richards Heuer’s study of analytic psychology: these weaknesses belong to how people think, so they do not disappear with effort or goodwill. That is why they repeat, and why the next primer deals with corrections that are built into the process, not left to willpower.

The Analyst’s Mind

Richards Heuer Jr. wrote Psychology of Intelligence Analysis for the CIA’s Center for the Study of Intelligence, which published it in 1999. It draws on cognitive psychology to explain why capable analysts reach flawed judgments.

Table of six biases documented by Heuer, from mind-set to hindsight, each with what goes wrong and the chapter of Psychology of Intelligence Analysis that covers it

Mind-Sets Filter What Is Seen

Heuer argues that perception is an active process of inference, not passive recording. Expectations shape what people notice and how they read it. A mind-set works like a lens, and he treats it as unavoidable, because nobody can process the volume of incoming information without simplifying assumptions. Mind-sets form quickly and resist change: new information tends to be absorbed into the existing picture, and early impressions can persist after clearer information arrives. His practical conclusion is that objectivity comes from making assumptions explicit so others can challenge them, not from having none.

Confirmation Comes Easily

Heuer describes analysts who accept the first explanation that seems to fit, a strategy he calls satisficing. He warns that an analyst trying to confirm a favored hypothesis is easily misled, because most supporting evidence is also consistent with other explanations. Evidence can make a hypothesis unlikely but rarely proves it impossible, and he concludes that the case for seeking disconfirming evidence is stronger than the case for seeking confirmation.

Evidence Is Weighed Unevenly

Heuer’s chapter on biases in evaluating evidence lists several habits. Vivid, first-hand or concrete information outweighs abstract or statistical data, even when the abstract data is sounder. Analysts rarely factor missing information into a judgment. Consistent evidence can mislead if it is redundant or comes from a small or biased sample. And beliefs formed from evidence often survive after the evidence is discredited, because the explanations people built around it remain in their heads.

Odds Are Judged by Recall and Starting Points

People judge how likely something is by how easily they can recall or imagine examples, so recent and vivid events seem more probable. Building a detailed scenario also makes it seem more probable, although each added step makes it less so. Estimates anchor on a starting number and are adjusted too little. Heuer also notes that words such as “possible” and “probable” mean different things to different readers, which is the problem Sherman Kent described and which we address in Confidence Tiers.

Governments Look More Coordinated Than They Are

Heuer notes that analysts tend to assume other governments act on coherent, rational plans. That overestimates their coordination and the predictability of what they will do next, and it overlooks blunders, bureaucratic bargaining and unintended consequences.

Hindsight Distorts the Review

The last of Heuer’s chapters on bias concerns hindsight. Knowing an outcome changes how people remember what they knew beforehand. Analysts overestimate how accurately they predicted events, readers underestimate how much a report taught them, and reviewers conducting a postmortem judge failures to have been more foreseeable than they were. Heuer reports that these biases persisted even when subjects were told about them and asked to compensate. This is the same caution the 9/11 Commission attached to its own report.

The Evidence

Better thinking does not fix thin or noisy evidence. Three problems recur.

Signals sit among noise. The 9/11 Commission cited Roberta Wohlstetter’s observation that it is much easier after an event to sort the relevant signals from the irrelevant ones. Before the event they arrive together, and the Commission said the warning system lacked methods used in earlier decades to detect surprise attacks.

More information is not always more accuracy. Heuer reports that once analysts have a minimum of necessary information, additional data usually raises their confidence more than their accuracy.

The adversary may be shaping what is seen. Cynthia Grabo, a US intelligence warning analyst, listed deliberate concealment and deception among the reasons warning fails, alongside mind-set and the problem of timing. Heuer adds that rejecting the possibility of deception too early biases the analysis, because analysts then stop looking for evidence that would support it.

The Organization

Individuals work inside institutions, and both reviews we use found failures at that level.

The 9/11 Commission found that information was not shared or pooled across agencies or across the foreign and domestic divide, and that no one clearly managed transnational cases. The WMD Commission found that the environment for Iraq analysis did not encourage skepticism about the prevailing view, and it described the community as not so much poorly managed as unmanaged. It also found that political pressure did not change analysts’ judgments, which is a reminder that the explanation does not have to be bad faith. Grabo adds a further organizational effect: repeated alerts that come to nothing breed skepticism, so a real warning may be discounted.

A Worked Example: Reading the Aviation Warning

This is an illustration, continuing the aviation example from the earlier primers. It describes how the habits above could shape a reading of such a warning. It is not a finding about any analyst or agency.

Three layers behind intelligence failure, the analyst's mind, the evidence and the organization, each with three causes drawn from Heuer, Grabo and the two commissions

A warning that echoes a well-known attack makes a hijacking scenario easy to picture, and an easily pictured scenario feels likely. That is availability at work. An analyst who expects a holiday-season warning to be routine may read every new detail as confirming routine, which is mind-set. The fact that officials have named no specific plot is a gap, and Heuer found that gaps rarely enter the judgment. A detailed story about how a plot could unfold feels convincing, although each added step makes it less likely. Each of these is a small, ordinary slip, and together they can move a judgment well away from the evidence.

Why the Failures Repeat

Three things keep these failures coming back. First, the biases belong to ordinary thinking, and Heuer reports that awareness of them does not reliably remove them. Second, the evidence will always be incomplete, so judgment will always be needed. Third, organizations change slowly, and the habits that produce failure, such as not sharing information or not rewarding doubt, are usually part of how they work. Each of these points to the same conclusion: protection has to be built into the process instead of relying on each analyst to resist their own mind.

How We Apply This

We treat our own judgments as exposed to the same weaknesses. Our confidence tiers separate what is sourced from what is judged, which makes gaps visible, and our structured analysis habits test more than one explanation. Our Foresight Record lets readers check our past calls, which works against hindsight. None of this makes us immune, and we say so.

Common Errors

  • Blaming the individual. Both reviews found failures built into process, structure and habit.
  • Assuming awareness is enough. Heuer reports that knowing about a bias does not reliably correct it.
  • Treating consistency as proof. Evidence that fits one story often fits others.
  • Ignoring what is missing. A gap in the evidence can itself be informative.
  • Reviewing with hindsight. Judging the past by what is known now makes failures look more foreseeable than they were.

Frequently Asked Questions

Why do intelligence analysts get things wrong?

Mostly for ordinary reasons: the mind filters and simplifies, the evidence is incomplete or noisy, and organizations shape what is asked, shared and challenged.

What is a mind-set?

A lens that filters perception and interpretation according to expectations. Heuer treats it as unavoidable and says the remedy is to make assumptions explicit so that others can challenge them.

Is more information always better?

No. Heuer reports that beyond a minimum, extra information usually raises analysts’ confidence more than their accuracy.

Does knowing about bias fix it?

Not reliably. Heuer reports that hindsight and anchoring biases persisted even when people were told about them, which is why protections are built into the process.

Can deception cause intelligence failures?

Yes. Grabo lists deliberate concealment and deception among the reasons warning fails.

What should I read next?

Correcting Intelligence Failures looks at the remedies the official reviews recommended and what they can and cannot fix.

Sources

How to Read Our Labels

Every claim in a CommandEleven Intelligence dossier carries one of these.

CONFIRMED

Multiple independent sources, direct corroboration, or reporting from a major established outlet. Stated as fact, with its source.

ASSESSED

Our reasoned judgment from the available evidence. Uncertainty remains.

REPORTED

One outside report we have not corroborated, from a source without established standing. Attributed, not vouched for.

ALLEGED

A party's claim about someone else's conduct. Attributed, not treated as established.

SPECULATIVE

Plausible but thinly supported. Context only, and never weighs on a conclusion.

DATA DEFICIT

Notation: the information does not exist or is out of reach. Stated, not guessed.

Read the full primer: Confidence Tiers, How We Label What We Know

Related dossiers

More from CommandEleven Intelligence

Opens in a new tab
Vol. I No. 1 is out
The monthly magazine

Convergence

Our dossiers, drawn together each month into one edition.

Get the first edition →
Publishing mid-December
The annual report

2027 Business Threat Exposure

Ten threats to American business, ranked by exposure.

Get notified →
The scorecard

Foresight Record

Every call we have made, scored in public against what happened.

Open →
The audit

GCTA Audit

The April 2026 forecasts, checked section by section.

Open →
The panel

Survival Dispatch News

Our regular panel appearances, with the full conversation.

Open →
The weekly show

Soft Targets America

Weekly on Tuesdays, with co-host Jason Piechura.

Open →
For organizations
Our core service

Intelligence as a Service

Private intelligence, delivered to your team.

Engage us →