IX Coach

After-Action Review: The US Army’s Tool for Continuous Learning

A four-question debrief that turns experience into usable learning

How do you run an after-action review?

An after-action review (AAR) is a structured debriefing developed by the US Army that asks four questions after any event: What was supposed to happen? What actually happened? What caused the difference? What do we do differently next time? Applied consistently, AARs build a culture of learning from experience rather than simply moving on. The evidence base is primarily observational and practice-based; experimental studies are limited but supportive.

The after-action review was developed by the US Army in the 1970s as a way to extract learning from training exercises before lessons were forgotten. The format is deliberately simple: four questions applied within hours or days of any significant event. Its power is not in the format but in the discipline — most individuals and teams move from event to event without closing the loop. The practices below apply the AAR framework to personal and professional contexts, with honest notes on what is established versus inferred.

Practices

How these practices connect

Every practice on this page exists to fix a specific way that unaided memory and self-perception distort learning from experience: fading recall, defensive self-protection, negative-event bias, evaporating intentions, and self-serving attribution all corrupt the causal story before a person can learn from it. The four-question format, timing, framing, and inclusion rules are each a targeted countermeasure to one of these distortions, not separate techniques competing for the same job.

Do I need to do all of these every time, or can I just ask the four questions?

The four-question AAR (what was planned, what happened, why the gap, what next) is the core structure and can be run alone for routine events. The other six practices are refinements that matter more as the stakes or blind-spot risk rises: separating judgment from analysis and choosing a next action apply to essentially every AAR, while bringing in a peer, running it on wins, and checking for system causes are worth adding specifically for higher-stakes or recurring situations. There is no rule requiring all seven every time; the four-question structure is the minimum viable version.

Why does it matter if I do the review right away versus this weekend?

Memory for specific situational details, such as the exact sequence of decisions or environmental conditions, fades within hours, while emotional impressions persist and often replace the facts with a narrative version of events. Waiting more than 48 hours means the analysis is being run on a retrospective story rather than the actual causal chain, which produces less accurate and less actionable conclusions. This is why the four-question AAR is specified as a within-24-hours practice rather than a weekly habit.

Why do I always end up blaming myself instead of figuring out what actually happened?

This happens because evaluative language like "that was a failure" or "I did badly" triggers defensive processing, which narrows the causal search toward self-protective explanations instead of accurate ones. The fix is to frame the analysis question as "what produced the difference between plan and outcome" rather than "what went wrong," which is the distinction between separating judgment from analysis and running a judgment-laden debrief. The same self-protective tendency also shows up as a bias toward blaming yourself specifically rather than the system you were operating in, which is a separate but related distortion worth checking for.

Should I only review the things that go badly, or the good stuff too?

Reviewing only setbacks produces a learning database skewed entirely toward what to avoid, because people are naturally more motivated to analyze failures than successes since negative events are more salient. Running the same four-question structure on a win identifies the specific conditions and decisions that caused it, which produces a template that can be deliberately recreated, rather than just a vague sense of having gotten lucky or being naturally good at something. A regular cadence that includes successes, not just crisis-triggered reviews, is what keeps the learning database from becoming one-sided.

I did the review and identified what went wrong, so why hasn't anything actually changed?

An AAR that produces insight without a concrete next step leaves the learning sitting as an aspiration in working memory, where it decays like any unconverted intention. Implementation intentions, if-then plans specifying exactly when and where a new behavior will occur, are well-established to dramatically increase follow-through compared to a general intention like "I need to prepare more." The fix is ending every AAR with something as specific as "before my next presentation, I will do a 90-minute dry run at least two days beforehand" rather than a general resolution.

Can I just review my own performance, or do I actually need someone else there?

Self-conducted AARs have a structural blind spot: self-serving attribution bias and narrative smoothing mean people unconsciously revise their account of events to protect self-esteem and preserve a coherent story, and this happens without the person being aware it is happening. A peer or coach who was not inside the event can ask questions that surface inconsistencies the reviewer cannot see from within their own narrative, but only if that person is willing to push back rather than simply validate the account. For high-stakes events this matters considerably more than for routine ones, where a solo four-question AAR is usually sufficient.

I keep having the same problem even though I keep fixing it after each review, what am I missing?

This pattern usually means the fix has been aimed at the individual level when the actual cause is a system-level one, such as an ambiguous process, missing information, or a misaligned incentive that will keep producing the same gap regardless of individual effort. Attribution research shows people default to personal explanations like "I made a mistake" over situational ones like "the process made this mistake easy to make," especially when analyzing their own performance. The corrective move is to explicitly ask, during the why-the-gap question, whether the system is reliably producing this outcome despite genuine individual effort, rather than repeating the same individual-level fix.

How do you choose between these?

For a routine, low-stakes event, the four-question structure run within 24 hours with a specific next action is enough on its own. As the stakes rise, or when the same type of problem keeps recurring despite prior fixes, add the system-versus-individual check to the why-the-gap question. When an event was significant enough that self-assessment is likely to be distorted, such as a major failure with reputational or emotional weight, bring in a trusted peer to facilitate rather than reviewing alone. Whether the event was a win or a setback should not change whether it gets reviewed; it changes only whether the causal analysis is aimed at replication or correction. Cadence-wise, the choice is between reviewing only when something goes wrong, which biases the entire learning process toward negative events, and reviewing on a regular schedule regardless of outcome, which is what actually converts accumulated experience into improved performance.

What these approaches do not settle

The peer-facilitation and success-AAR practices both rest on evidence gathered from team debriefs and general research on analyzing outcomes, extrapolated to a single person reviewing their own performance alone; the underlying mechanisms, such as self-serving attribution bias and salience-driven negativity bias, are well documented, but whether an individual doing this solo captures the same benefit as a facilitated team debrief is not directly tested. The strongest evidence in this set is for implementation intentions themselves, which come from randomized controlled trials, but the specific step of converting an AAR insight into that if-then format is a practitioner-derived combination rather than something tested as a unit. The system-versus-individual distinction draws on safety engineering research from high-consequence domains like aviation and medicine, where the cost of missing a system cause is severe; applying the same distinction to ordinary personal or work performance is a reasonable generalization of the mechanism, but the stakes and incentive structures are not the same, so the size of the benefit likely differs.

Practice this with IX Coach

Practice this with IX Coach

Frequently asked questions

How do you run an after-action review?

An after-action review (AAR) is a structured debriefing developed by the US Army that asks four questions after any event: What was supposed to happen? What actually happened? What caused the difference? What do we do differently next time? Applied consistently, AARs build a culture of learning from experience rather than simply moving on. The evidence base is primarily observational and practice-based; experimental studies are limited but supportive.

How do you run the four-question AAR within 24 hours of any significant event?

Memory for specific situational details degrades rapidly after events; encoding fades within hours while emotional impressions persist longer. Running an AAR within 24 hours captures specific facts — decisions, sequences, environmental conditions — that will not be available at a weekly review. The structured format redirects attention from blame to causation, which is what produces actionable learning rather than morale damage.

How do you separate judgment from analysis in the debrief?

Evaluative language ("that was a failure," "I did badly") activates defensive processing, which narrows the causal search to self-protective explanations rather than accurate ones. The Army’s format is explicitly non-judgmental: the question is "what caused the gap," not "who was at fault." The separation allows causal accuracy, which is the precondition for actionable improvement.

How do you apply the AAR to successes, not only setbacks?

Humans are more motivated to analyse failures than successes because negative events are more salient. This systematic bias means the learning database is skewed toward what to avoid rather than what to replicate. An AAR on a successful event asks the same four questions: identifying the specific conditions and decisions that contributed to the outcome produces a template that can be deliberately recreated rather than hoped for.

How do you end every AAR with one specific, time-bound next action?

Implementation intentions — if-then plans specifying when and where a new behaviour will be performed — dramatically increase follow-through on intentions. An AAR insight that is not converted into an implementation intention exists as an aspiration in working memory, where it decays. A specific "next time X happens, I will do Y" converts the learning into a cued response that fires automatically.

How do you build a regular AAR cadence, not just crisis-triggered reviews?

Deliberate practice theory holds that improvement is driven by systematic, feedback-rich engagement with performance — not by accumulated time. A regular review cadence provides the feedback loop that makes experience into expertise; absence of regular feedback means experience simply accumulates without converting to skill. Crisis-only reviews bias the learning database toward what went badly, missing the positive feedback that builds confidence and reinforces effective practices.

How do you include a trusted other in high-stakes AARs?

Self-serving attribution bias and narrative smoothing are well-documented in memory and attribution research: people unconsciously revise their account of events to protect self-esteem and preserve a coherent narrative. A facilitating partner who was not inside the event can ask questions that surface inconsistencies in the causal account that the reviewer, within their own narrative, cannot access.

How do you distinguish system causes from individual causes in the analysis?

Attribution research shows humans default to personal attribution ("I made a mistake") over situational attribution ("the system made this mistake easy to make"), especially in self-analysis. System causes — ambiguous processes, missing information, misaligned incentives — tend to recur until addressed structurally. Individual-level fixes on system problems produce temporary improvement followed by relapse.

Do I need to do all of these every time, or can I just ask the four questions?

The four-question AAR (what was planned, what happened, why the gap, what next) is the core structure and can be run alone for routine events. The other six practices are refinements that matter more as the stakes or blind-spot risk rises: separating judgment from analysis and choosing a next action apply to essentially every AAR, while bringing in a peer, running it on wins, and checking for system causes are worth adding specifically for higher-stakes or recurring situations. There is no rule requiring all seven every time; the four-question structure is the minimum viable version.

Why does it matter if I do the review right away versus this weekend?

Memory for specific situational details, such as the exact sequence of decisions or environmental conditions, fades within hours, while emotional impressions persist and often replace the facts with a narrative version of events. Waiting more than 48 hours means the analysis is being run on a retrospective story rather than the actual causal chain, which produces less accurate and less actionable conclusions. This is why the four-question AAR is specified as a within-24-hours practice rather than a weekly habit.

Why do I always end up blaming myself instead of figuring out what actually happened?

This happens because evaluative language like "that was a failure" or "I did badly" triggers defensive processing, which narrows the causal search toward self-protective explanations instead of accurate ones. The fix is to frame the analysis question as "what produced the difference between plan and outcome" rather than "what went wrong," which is the distinction between separating judgment from analysis and running a judgment-laden debrief. The same self-protective tendency also shows up as a bias toward blaming yourself specifically rather than the system you were operating in, which is a separate but related distortion worth checking for.

Should I only review the things that go badly, or the good stuff too?

Reviewing only setbacks produces a learning database skewed entirely toward what to avoid, because people are naturally more motivated to analyze failures than successes since negative events are more salient. Running the same four-question structure on a win identifies the specific conditions and decisions that caused it, which produces a template that can be deliberately recreated, rather than just a vague sense of having gotten lucky or being naturally good at something. A regular cadence that includes successes, not just crisis-triggered reviews, is what keeps the learning database from becoming one-sided.

I did the review and identified what went wrong, so why hasn't anything actually changed?

An AAR that produces insight without a concrete next step leaves the learning sitting as an aspiration in working memory, where it decays like any unconverted intention. Implementation intentions, if-then plans specifying exactly when and where a new behavior will occur, are well-established to dramatically increase follow-through compared to a general intention like "I need to prepare more." The fix is ending every AAR with something as specific as "before my next presentation, I will do a 90-minute dry run at least two days beforehand" rather than a general resolution.

Can I just review my own performance, or do I actually need someone else there?

Self-conducted AARs have a structural blind spot: self-serving attribution bias and narrative smoothing mean people unconsciously revise their account of events to protect self-esteem and preserve a coherent story, and this happens without the person being aware it is happening. A peer or coach who was not inside the event can ask questions that surface inconsistencies the reviewer cannot see from within their own narrative, but only if that person is willing to push back rather than simply validate the account. For high-stakes events this matters considerably more than for routine ones, where a solo four-question AAR is usually sufficient.

I keep having the same problem even though I keep fixing it after each review, what am I missing?

This pattern usually means the fix has been aimed at the individual level when the actual cause is a system-level one, such as an ambiguous process, missing information, or a misaligned incentive that will keep producing the same gap regardless of individual effort. Attribution research shows people default to personal explanations like "I made a mistake" over situational ones like "the process made this mistake easy to make," especially when analyzing their own performance. The corrective move is to explicitly ask, during the why-the-gap question, whether the system is reliably producing this outcome despite genuine individual effort, rather than repeating the same individual-level fix.

Research

Related concepts

Articles that cover this

How this pairs with other practices