How to Debrief a Training Exercise: A Three-Phase Guide That Actually Changes Behavior

Our pilot sessions for Operation Aetherfall run about seven and a half hours. Four of those hours are gameplay – six players, a commander named Korr working from a flawed plan, secret information cards, and a dungeon that punishes bad decisions. The part I plan most carefully is the ninety minutes after the dice are put away.

To debrief a training exercise, run three phases in order: a reactions phase where participants decompress emotionally (10-15 minutes), an analysis phase where you examine specific moments from the exercise and the reasoning behind them (40-50 minutes), and an application phase where every participant writes a specific if-then commitment to change one behavior at work (25-30 minutes). Each phase does a job the others can't. Skip reactions and people analyze poorly because they're still keyed up. Skip application and the insights evaporate before Monday.

That structure isn't mine. It shows up in nearly every evidence-based debriefing framework across medicine, the military, and corporate training. What I can add is what it looks like at an actual table, with real timing budgets, real questions, and moments from our pilots.

Why does the debrief matter more than the exercise?

Learning from experience mostly happens in the structured conversation afterward, and the effect sizes are unusually large for training research. A meta-analysis by Tannenbaum and Cerasoli (2013) found that properly conducted debriefs improve individual and team performance by roughly 20-25% (d = .67), with effects consistent across medical, military, and organizational settings. A second meta-analysis focused on after-action reviews found a larger effect still – d = 0.79 across 61 studies covering 915 teams and 3,499 individuals.

Those are remarkable numbers for a conversation.

The game-based learning research explains why. Some learning happens during play, but the deeper lessons – and nearly all of the transfer to workplace behavior – get drawn out during the debrief. Participants have the experience; they don't automatically have the meaning. This is a big part of why workplace training doesn't stick: the exercise was vivid, the meaning-making got skipped, and three weeks later there's a fond memory and no changed behavior.

When I tell people we budget ninety minutes of talking after a four-hour game, some of them wince. The research says the wince points the wrong way.

What are the three phases of a training debrief?

Every major debriefing framework converges on the same three-phase shape: reactions, then analysis, then application. Fanning and Gaba named the phases Reaction, Analysis, and Summary in their 2007 work on simulation debriefing. The GAS model from UPMC's WISER simulation center calls them Gather, Analyze, and Summarize, and recommends spending about 25% of debrief time on the first, 50% on the second, and 25% on the third. The military's after-action review asks four questions in the same sequence: What was supposed to happen? What actually happened? Why was there a difference? What can we do differently next time? Thiagarajan's six-phase model for learning games slices it finer – How do you feel? What happened? What did you learn? How does this relate? What if? What next? – but the progression is identical.

When army officers, anesthesiologists, and corporate trainers independently land on the same structure, I stop treating it as a preference and start treating it as how brains convert experience into learning.

Here's our budget for a 90-minute debrief following roughly four hours of gameplay:

Phase	Time
Phase 1: Reactions	10-15 minutes
Transition: individual written reflection	5 minutes
Phase 2: Analysis	40-50 minutes
Phase 3: Application and commitment	25-30 minutes

The research is flexible on total duration. Twenty to thirty minutes is most common for shorter simulations, and the current healthcare simulation standards explicitly warn against formulas tied to exercise length – complexity of objectives, group composition, and learner experience should set the number. Quality of reflection beats minutes on the clock.

(The four hours of play that feed this debrief are their own craft. I've written separately about how to run a tabletop RPG training session.)

How do you run the reactions phase?

Start by letting people say how they feel, and resist every urge to make it productive. This phase exists for emotional decompression before intellectual analysis. Participants who are still activated from the exercise can't examine their own behavior objectively yet. The energy has to settle.

We actually start settling it before the debrief begins: when gameplay ends, we take a ten-minute break. Stretch, water, casual conversation. Rolling straight from the exercise into the debrief feels efficient and works worse.

Then three questions, asked to the room:

• "How are you feeling right now?"
• "What was the most intense moment for you?"
• "What surprised you about yourself during the game?"

Someone will start analyzing immediately – "I think I should have pushed back earlier when..." Gently redirect: "We'll dig into that in a minute. Right now, just – how are you feeling?" The analysis is coming. It'll be better for the wait.

Before opening group discussion, we add five minutes of private writing. Each participant gets a card with three prompts: one moment where I handled a conflict well, one moment I wish I'd done differently, something I noticed about my conflict patterns. Nobody else reads it. Studies of nursing simulations found that students specifically valued a self-debrief period before group discussion – it gave them room to organize their thoughts without social pressure. At our table it does something else too: it keeps the loudest voices from setting the frame before quieter participants have formed their own.

How do you run the analysis phase?

The analysis phase lives or dies on specificity. Generic questions produce generic answers. "What did you learn about conflict today?" gets shrugs and platitudes. "Walk me through that moment when Bramm pushed back on Korr's timeline" gets real analysis.

That specificity has to be gathered during the exercise. While the game runs, I keep a notepad and jot down three to five moments: a question that changed a conversation's direction, someone challenging the commander (or visibly deciding not to), how the group actually reached a decision, someone holding back when they had something to offer, a disagreement starting to feel tense. Those notes are the raw material for the entire analysis phase.

The second tool is a questioning technique called advocacy-inquiry, from Rudolph and colleagues' "Debriefing with Good Judgment" work at Harvard's Center for Medical Simulation. Pair an honest observation with a genuinely curious question about the person's reasoning: "I noticed you chose to address Korr by leading with the structural data rather than directly challenging his plan. Help me understand what was driving that decision." The idea underneath it: people's actions are driven by their frames – their knowledge, assumptions, and feelings – and you can't reframe what you haven't surfaced. Advocacy-inquiry surfaces reasoning without making anyone feel judged, which is most of the trick of a good debrief.

One more finding worth taping inside your notepad: debriefs where the facilitator talked less than the participants were rated notably higher in quality than debriefs the facilitator dominated. Your job is questions and silence. If you're explaining, you're probably in the way.

Here's a bank of analysis-phase questions from our pilot sessions, grouped by what they surface:

Positions and interests

• "When one of you wanted X and another wanted Y, what was underneath those positions? What did each of you actually need?"
• "When did the group get stuck arguing about solutions instead of understanding the problem?"
• "Was there a moment where someone asked a question that changed the direction of the conversation? What made it effective?"

Challenging authority

• "How did you handle the disagreement with the commander? What did you try first?"
• "Can you put words to what made a challenge feel constructive rather than insubordinate?"
• "What stopped the group from pushing back harder? What would have made it easier?"

How decisions actually got made

• "How did the group make that decision? Walk me through it."
• "Whose input was included? Whose wasn't? Why?"
• "Did anyone feel like a decision was made before everyone had been heard?"

What stayed unsaid

• "Was there a moment you held back from saying something? What stopped you?"
• "What did it cost the team when important information stayed hidden?"

Escalation

• "Did any disagreement start to feel personal? When did you notice that shift?"
• "What would have helped de-escalate that moment?"

A note on the "held back" questions: normalize without pathologizing. Most people hold back in situations like these – it's a very human response. The question worth asking isn't whether someone avoided. It's what the avoidance cost.

How do you connect game events to workplace behavior?

The bridge from exercise to workplace has to be built deliberately, because transfer doesn't happen on its own. The simulation-game research identifies the building blocks: pre-brief framing that establishes the metaphor before play begins, in-game moments where the facilitator notes parallels, and debrief questions that ask directly how what happened in the game resembles what happens on the participant's real team.

We plant the bridge before the game starts. Part of our pre-game framing: "During the game, you're going to face decisions under pressure with incomplete information, competing priorities, and authority figures who may not have the full picture. Sound familiar?"

My favorite facilitation assist in the whole session is a villain. One of Operation Aetherfall's adversaries, Overseer Vane, has an ability that forces a player to mark Stress if they can't justify their last action to their teammates. During play it's a game mechanic with pointed taunts – "Do your teammates know why you just did that? Because I'm curious." In the debrief it becomes two questions: "When Vane forced you to justify your actions to each other, what was that like?" and then, "How often at work do you make a decision without explaining your reasoning to the people affected by it?" That second question is the whole product in miniature. The villain does the teaching. The debrief makes it land.

Direct bridging questions close out the analysis phase: "Where have you seen this exact dynamic at your job?" "What made this conflict feel familiar?" And the one that opens the most interesting conversations: "What did the game make possible that your work environment doesn't?"

How should a debrief end?

End with written, specific, spoken-aloud commitments. The research is blunt here: debriefs that skip an explicit application phase leave transfer potential on the table. Thiagarajan's model dedicates its final two phases to it ("What if?" and "What next?"), and the corporate training literature treats the commitment phase as doing double duty – imprinting the learning and creating accountability.

We use implementation intentions: if-then plans anchored to a specific cue. Each participant fills out a card that reads:

When I notice: ________ (a feeling, a situation, a pattern you recognized today)

Instead of: ________ (my usual default response)

I will: ________ (a specific action – something I practiced or learned today)

And if it doesn't go well or I forget, I will: ________

The specificity is the mechanism. "If there's a conflict, I'll handle it better" is a wish. "If I notice myself shutting down when my manager pushes back on my estimate, I will say 'Can you help me understand what's driving the timeline?'" is a plan. The research behind this deserves its own article – it has one: implementation intentions and training transfer – but the headline number is that specific if-then plans roughly double follow-through, from about 30% to about 67%.

Then each participant reads their plan to the group. Sharing a commitment publicly raises follow-through from roughly 43% to 76%, and the group helps sharpen each plan as it's read: "What's the internal signal that tells you this situation is happening?" "What words would you actually say?"

Last, accountability partners. Everyone picks one person from the room, and in two weeks that person checks in with a single question: "How's your plan going?" Not a long conversation. Just a check-in.

We close by re-running the same assessment participants took before the session, which is how we measure whether anything actually shifted. The design of those instruments is covered in pre- and post-training surveys.

Where do debriefs go wrong?

The failure modes are consistent enough to list:

• Skipping the reactions phase. Analysis starts while people are still activated, so the conversation stays defensive or shallow. The decompression isn't a warm-up. It's a precondition.
• The facilitator talks too much. When facilitator talk time exceeds participant talk time, measured quality drops. Ask, then wait.
• Generic questions. "What did you learn?" produces the answers people think you want. Point at specific moments you noted during the exercise.
• Ending on insight instead of commitment. "That was eye-opening" is where debriefs go to die. No written if-then plan, no transfer.
• Letting the debrief absorb schedule slippage. When the exercise runs long, the debrief gets compressed because it looks like the optional part. Given that the debrief carries most of the learning, that's exactly backwards. Protect the ninety minutes the way you'd protect the game itself.
• The marathon debrief. Cognitive overload is a real risk past about 45 minutes of unbroken discussion. Our 90 minutes works because it's segmented – reactions, written reflection, analysis, application – with a change of activity at each seam.

The exercise makes the experience. The debrief makes the learning. We spend four hours playing Operation Aetherfall and ninety minutes talking about what happened – and if I ever had to cut one of them short, it wouldn't be the talking.