After-Action Review

Thinking Tools6.4 Conducting Follow-Up and Taking FeedbackAlso known as AAR

From United States Army training doctrine, as described by Gordon R. Sullivan and Michael V. Harper (Hope Is Not a Method, 1996).

A short team review held right after an action: what was meant to happen, what did, why the two differed, and what to change, with rank left at the door.

Four questions after a launch that slipped, and the changes they produced

Play the review, or select a question to read the team's answer. The plan sits beside what happened, the gap gets its causes, and the changes feed the next launch's plan. Illustrative.

  • Into the next launch's plan (What to change to Meant to happen)
  • Facts disputed: rebuild the log (Why they differed to What happened)

Select a stage to read it, or play the sequence.Meant to happen. The plan as it stood: invoicing released to all 4,000 accounts on Tuesday at nine, help articles live, support briefed on Monday, and 300 accounts using it in the first week. Write the plan down before anyone explains anything.What happened. The release slipped to Thursday when the migration check failed. The help articles went live; support was never briefed and took 210 tickets in two days. 120 accounts used it in the first week.Why they differed. The migration check ran on staging data three months old. The support briefing was tied to the Tuesday date; when the check failed on Monday it was called off with the release and never rebooked, because the go call was the product lead's alone and support was not on it.What to change. Run the migration check on a production copy two days out (the platform lead, from the next release). Book the support briefing for the hour after the go call, and put the support lead on that call (the product lead, this week).
  1. Meant to happen. The plan as it stood: invoicing released to all 4,000 accounts on Tuesday at nine, help articles live, support briefed on Monday, and 300 accounts using it in the first week. Write the plan down before anyone explains anything.
  2. What happened. The release slipped to Thursday when the migration check failed. The help articles went live; support was never briefed and took 210 tickets in two days. 120 accounts used it in the first week.
  3. Why they differed. The migration check ran on staging data three months old. The support briefing was tied to the Tuesday date; when the check failed on Monday it was called off with the release and never rebooked, because the go call was the product lead's alone and support was not on it.
  4. What to change. Run the migration check on a production copy two days out (the platform lead, from the next release). Book the support briefing for the hour after the go call, and put the support lead on that call (the product lead, this week).

An example

A software company released its new invoicing feature two days late, and support took 210 tickets in the first two days. The product lead called a review for the next morning with the engineers, the two support agents who had taken most of the calls, and the platform lead. The head of product sat in and spoke last.

They wrote the plan on one side of the whiteboard: a Tuesday release to all 4,000 accounts, support briefed on Monday, 300 accounts using it in week one. On the other side went what happened: a Thursday release after a failed migration check, no briefing, 120 accounts using it. A support agent pointed out that nobody had told them the date had moved. The causes took twenty minutes: test data three months old, and a briefing tied to a date when it should have been tied to a decision. They left with two changes, each with an owner and a start date, and the next launch plan listed both on its first page.

Why it works

The review works because it compares two records instead of trading opinions. Writing down what was meant to happen before discussing what did fixes the standard, so the gap is a fact on the board and nobody can quietly redefine success afterward. Separating what happened from why it happened keeps the timeline from being told as a defense.

The ground rules carry the rest. Held within a day, the review catches details that memory smooths away within a week. Leaving rank at the door lets the people closest to the work, here the support agents, say what the senior people did not see; the missing briefing came from them. And because the last question asks for specific changes with owners, the meeting ends with something the next plan can use, which is the only test of whether the review was worth holding.

The thinking behind it

Gordon R. Sullivan Gordon R. Sullivan was the 32nd Chief of Staff of the US Army and led it through its downsizing and transformation after the Cold War. The after-action review comes from Army training doctrine: the routine of reviewing an exercise or an operation with the people who took part in it. Sullivan, with Michael Harper, described it for business readers in Hope Is Not a Method (1996), their account of how the Army changed and kept learning while it shrank, told by the general who had led that change.

Michael V. Harper Michael V. Harper wrote Hope Is Not a Method (1996) with Sullivan: an account, written for business leaders, of how the US Army changed and learned as it shrank after the Cold War. Its title states the theme in four words: hoping the next effort goes better is no way of making it go better. The after-action review is one of the Army practices the book describes for those readers, as a habit an organization facing large change can take up for itself.

How to run it

  1. Hold it within a day. Book the review before the action ends where possible, while people still remember the order of events.
  2. Bring the people who acted. Invite everyone who did the work and anyone who watched it closely, including the most junior.
  3. Set the ground rules aloud. No rank, no blame, specifics only; the most senior person speaks last.
  4. Write down what was meant to happen. Record the plan, the dates, and the target numbers as they stood beforehand, before any explanation.
  5. Write down what happened. Build the timeline from logs, tickets, and messages, and settle disputed facts before moving on.
  6. Explain the gap. For each difference, ask what caused it until the cause is something the team could change.
  7. Agree specific changes with owners. Write each change as an action with a name and a date, and carry it into the next plan.

Where it goes wrong

The review turns into a blame session when the most senior person speaks first or reacts badly to criticism. People then explain why the gap was not their fault, and the timeline bends to fit. Watch who talks; if the junior people go quiet, the rules have stopped holding, and the senior person has to say so and invite them back in.

Lessons also come out too vague to use. "Communicate better" changes nothing; "brief support in the hour after the go call" does. Push every lesson until it names an action, an owner, and a date.

The quietest failure is lessons that never reach the next plan. If the same cause appears in three reviews, nobody is reading them. Put the previous review's changes at the top of the next plan.

When to use something else

  • Blameless Postmortem. When an incident hurt customers and needs a written analysis of its system causes that other teams will read.
  • Retrospective Formats. When a team reviews a recurring sprint or cycle of work, without one action to examine.
  • Five Whys. When one gap in the review needs its cause traced several levels down before a fix is chosen.

Where it sits in SOLVE eX

Go to the source

  • Hope Is Not a Method, 1996

Back to the catalog