Article

Reviewing a finished piece of work: the debrief, and whether it changes anything

A debrief is a structured review of work already finished, held to change how the next piece goes. The evidence for holding one is strong: across 61 studies covering 3,499 people, debriefing improved the measured outcomes by d = 0.79. Alignment and an objective record are what carry it. This page is for the review you would hold on your own work, or with the few people who did it with you. It covers what the evidence actually measured, which parts of a debrief produce the effect, and where military and clinical studies stop telling you about your project. Holding fewer meetings in the first place belongs to the meetings pillar.

What does a debrief actually produce?

The project shipped on Thursday. By Monday two of the people who did it have been moved onto something else, and the hour you blocked to talk about how it went is the easiest hour in the week to hand back. You have given it back before.

Keiser and Arthur put the largest number on that hour in the Journal of Applied Psychology in 2021. They pooled 61 studies of the after-action review, which is the military and clinical name for a debrief, reporting 107 effect sizes across 915 teams and 3,499 individuals. The overall improvement came out at d = 0.79 on the evaluation criteria those studies measured, which the authors note sits above some of the largest effects previously reported for training methods in general. That is a large effect for something you can hold in an hour.

That figure pools multiple evaluation criteria rather than tracking one outcome. So it supports a general claim that reviewing finished work helps, and it does not predict what will move on the single thing you care about in your own project.

A second pool had already pointed the same way with a smaller number. Tannenbaum and Cerasoli pooled 46 samples covering 2,136 people for Human Factors in 2013, and found debriefs improved effectiveness over a control group by roughly 25 percent, or d = 0.67. Two independent reviews, eight years apart, disagreeing only about size.

Both pools draw mostly on the military and on healthcare, where a team performs, gets reviewed, and performs again with a score attached. That sounds like a reason to discount the whole literature for your project. Tannenbaum and Cerasoli tested it directly: average effects came out similar for teams and for individuals, in simulated and in real settings, and in medical and non-medical samples alike. The format travels further than its sample list suggests.

Except the second run.

An effect size needs a later performance to measure, so almost every study in both pools reviewed people who went on to do the task again. A piece of work that happens once, by a group that disbands afterwards, is the case this evidence is quietest about. Your project may be exactly that. Read d = 0.79 as a strong argument for reviewing work you repeat, and as a weaker one, though still the best going, for work you do not.

Which part of the review is doing the work?

Both meta-analyses went looking for that, which is what makes them useful to you rather than merely encouraging. Each tested which features of a review predicted the benefit, and they converged on a short list.

Alignment came out of both. Keiser and Arthur found it one of only two characteristics that consistently contributed to how well a debrief worked. Tannenbaum and Cerasoli found the same bolstering effect, and spelled out what it covers: the people in the room, the focus and intent of the review, and the level you are measuring all have to match. Held with the team that did the handover, about the handover, measuring the handover, a review is aligned. Held with whoever was free on Friday, about the project in general, it is a conversation you will enjoy and not act on.

That distinction is not pedantry.

It decides whether your review changes anything.

The other consistent characteristic was objective performance review media, which means the review runs better on a record of what happened than on what the room remembers. Keiser returned to that question in 2024 with a systematic review of 91 studies covering 113 after-action reviews, asking what technology contributes. Most of it turned out to sit either in the task itself, in 92 percent of those reviews, or in delivering feedback afterwards, in 52 percent, and the review found little guidance available on using any of it well. The useful part for you is smaller than the technology question.

The modest version of an objective record is already in your tools: the ticket history, the dated plan beside what actually shipped, the thread where the scope changed. You need no new software. All of it outranks five people trying to remember March. In practice the gap shows up within minutes, when two of them describe the same slipped deadline and neither account matches the dates. Pull the dates first.

Structure is where the evidence declines to hand you a template. A highly structured review beat a loose one in the military studies, while in healthcare high and low structure performed about the same. Keiser and Arthur read the whole picture as characteristics that interact rather than a recipe that stacks, and concluded that the effectiveness of a debrief has to be understood as a combination of them. No single correct shape comes out of the research, which is awkward for anybody selling one. So your shape has to be earned rather than copied.

Can you run one yourself, without a facilitator?

Yes. You can run it yourself, and on this evidence that is not a compromise.

Keiser and Arthur tested facilitation in combination with the other characteristics rather than on its own, and the two strongest combinations they report are both self-led: self-led with a team-aligned debrief, and self-led with objective review media. An outside expert running the session was not the condition that produced the largest effects.

That matters, because the missing facilitator is the usual reason the review never happens. Nobody owns it, nobody has been trained to run it, and the slot quietly comes back. What the research supports is narrower and more useful than a trained chair. You can review your own work, provided you hold the review at the level you can act at, and open the actual record while you do.

Neither pool reports how long a debrief should take, so a specific number from anybody is not coming from this research. Nobody has measured the length. Tannenbaum and Cerasoli do describe debriefs as a relatively inexpensive and quick intervention, and put the gain from properly conducted ones in the region of 20 to 25 percent. Short enough that you will hold it again is the operative constraint. Timeboxing is the mechanism when a review of one piece of work keeps expanding into a discussion of the last six months.

Who this does not suit is worth saying plainly. A self-led review assumes you can write down what went wrong without it being read back to you later. In some teams it does not hold. Where it does not, the honest version is a private review that improves your own next run, rather than a group session producing a diplomatic document. It also assumes a next run exists. Work that will never repeat, by a group already dissolved, gets less from this than the headline numbers imply.

If the reviews are worth keeping, they need a slot rather than good intentions. Getting Things Done builds its weekly review in as a fixed appointment, and the same logic holds at project scale: the review that happens is the one with a time on your calendar.

How does your working style change the answer?

Which part of this you get wrong tends to follow how you already work. The four names below are this site's own, from the working-style self-check on the home page.

Architect-leaning readers build the template first, and the structure finding is the one to read twice: more structure helped in the military studies and made no measurable difference in the clinical ones. Your form is not the active ingredient. Spend the preparation on pulling the record together instead, because that one is.

Sprinters skip the review because the work shipped and the next thing is already moving. The fix is timing rather than discipline. Hold it the same day, while the dates and the thread are still open in front of you, in one short block you do not reschedule.

Visionary-leaning readers turn a review of one project into a review of how everything is going, which is the alignment failure both meta-analyses name, in its most enjoyable form. Decide the level before you start and say it out loud. Anything bigger than that level is a separate conversation, and it will survive being parked.

Improvisers run the whole thing from memory on the way back from lunch, get something real out of it, and that is why the habit survives. The objective-media finding is aimed squarely here: five minutes with the actual ticket history first changes what you conclude, and costs you nothing you were already using.

So take the last piece of work you finished. Open its record, decide whether you are reviewing your own part or the team's, and hold a short self-led review at that level this week. Whether it then becomes a standing slot on everybody else's calendar is a different question, and the meetings pillar covers the case for adding one.

Frequently asked questions

Do project debriefs actually work?

Yes, and the evidence is unusually strong: a meta-analysis of 61 studies covering 3,499 people found an overall improvement of d = 0.79. An earlier pool of 46 samples put the gain at about 25 percent, or d = 0.67. Most of those studies ran in the military and in healthcare, with people who went on to do the task again, so the figures are firmest for work you repeat.

Can you run a debrief without a facilitator?

Yes, and the two strongest combinations in the 2021 meta-analysis were both self-led. Self-led paired with a team-aligned review, and self-led paired with an objective record of what happened, produced the largest effects. An outside expert was not what maximised the benefit, so a missing facilitator is not a reason to skip the review.

How long should a debrief meeting take?

Neither meta-analysis reports a duration, so a specific number from anybody is not coming from this research. What the 2013 review does say is that debriefs are a relatively inexpensive and quick intervention. Treat the length as whatever you will hold and repeat, and spend what you have on the record rather than on the agenda.

Why do lessons-learned meetings never change anything?

Most often because the review is not aligned: the people in the room, the work being reviewed, and the level anybody can act at do not match. Alignment is one of only two characteristics that consistently predicted benefit across 61 studies. A session held with whoever was available, about the project in general, leaves no level to act at. What happens to the actions afterwards is a separate problem with its own page on this site.