The Debrief Is Too Late to Fix What Was Never Agreed

I have facilitated debriefs where the hardest question was the first one, and it was not hard because the event went badly. It was hard because nobody in the room could answer it.

The question is: what was supposed to happen?

When six people give six different answers, the review stops being a review of the event and becomes an archaeology of the assumptions. That is still useful. It is just far more expensive than settling the question in a ninety-minute meeting before anyone showed up.

This is why I treat planning alignment and observation as a distinct engagement rather than as a courtesy call before the debrief. Most of what an After Action Review uncovers was decided, or left undecided, weeks earlier.

Role ambiguity is not a soft problem

The phrase people use is “everybody knows what they are doing.” The research does not support that confidence.

Tubré and Collins meta-analyzed the relationship between role ambiguity and job performance and found a negative relationship of ρ = −.21, with their conclusion stated directly: role ambiguity ought not to be dismissed as an unimportant variable in the job performance domain.¹³ The same analysis found role conflict had a negligible relationship to performance at ρ = −.07, which is a distinction worth holding onto. Competing demands are survivable. Not knowing what you own is not.

The earlier Jackson and Schuler meta-analysis shows what ambiguity costs beyond output. Across tens of thousands of participants, role ambiguity correlated with tension and anxiety at .47, with propensity to leave at .29, and with organizational commitment at −.41.⁶ An organization that never writes down who owns what does not simply perform worse. It burns the people who keep guessing.

Alignment is measurable, and it predicts performance

The technical term for what a committee has when it is actually aligned is a shared mental model: organized mental representations of the key elements in a team’s environment that are held in common across members.¹¹

Mathieu and colleagues studied this directly with dyads flying a combat flight simulation and found that both shared task-based and shared team-based mental models related positively to subsequent team process and performance. The important part is the mechanism: team processes fully mediated the relationship between mental model convergence and team effectiveness.¹⁰ Alignment does not make people work harder. It makes the handoffs work.

DeChurch and Mesmer-Magnus cumulated 231 correlations from 65 independent studies and found team cognition related to team performance at ρ = .38, to team behavioral process at ρ = .43, and to motivational states at ρ = .37. It explained a significant additional 6.8 percent of variance in team performance beyond what behavioral process and cohesion already accounted for.²

In plain terms: a team that likes each other and communicates well still underperforms a team that also agrees on what the job is.

Define success before, not after

The second thing I insist on in a planning session is a written definition of what success looks like. This is the part committees resist, because a vague goal cannot be missed.

Locke and Latham summarized thirty-five years of goal-setting research and reported that specific, difficult goals consistently produced higher performance than urging people to do their best, with effect sizes across meta-analyses ranging from .42 to .80. Their line is blunter than anything I would write: when people are asked to do their best, they do not do so.⁹

A performance agreement written before the event converts “let us make this a great conference” into something that can be evaluated afterward without argument.

The briefing itself is the intervention

There is a body of evidence that a short structured briefing before the work changes the work, and the interventions studied are almost embarrassingly small.

Lingard and colleagues ran a thirteen-month study in a Canadian academic hospital, with trained observers scoring 172 procedures against a validated scale. Communication failures per procedure fell from a mean of 3.95 before the briefing intervention to 1.31 after. In thirty-four percent of briefings, the briefing itself surfaced a problem, closed a critical knowledge gap, or produced a decision.⁸

Nundy and colleagues studied a two-minute standardized discussion led by the attending surgeon before each procedure. Unexpected delays fell by 31 percent overall and by 82 percent among surgeons, with a 19 percent reduction in communication breakdowns leading to delays.¹²

The best-known example is the World Health Organization surgical safety checklist. Across eight hospitals on four continents and roughly 7,700 patients, the death rate was 1.5 percent before the checklist and 0.8 percent after, and inpatient complications fell from 11.0 percent to 7.0 percent.⁵ I want to be precise about what that study is and is not. It is a pre-intervention and post-intervention study, not a randomized trial, and the correct verb is that the checklist was associated with those reductions. A later systematic review reached risk ratios of 0.59 for any complication and 0.77 for mortality while stating plainly that the evidence cannot be regarded as definitive.¹ That is still a remarkable return on a nineteen-item list and two minutes of everyone’s attention.

Why I ask to observe

The third element is the one organizations find intrusive, and it is the one I will not give up.

Self-reported performance is not evidence. In one study of hand hygiene, healthcare workers reported complying before touching a patient 76.9 percent of the time. Directly observed compliance for the same moment of care was 11.7 percent, thirty of 257 observed instances.⁴ That is a gap of roughly sixty-five percentage points, and nobody in that study was lying. They believed their answer.

Observation is better. It is not perfect, and I will say so rather than sell it as perfect. When researchers compared directly observed hand hygiene against automated electronic monitoring, staff performed eight hygiene events per hour when unobserved and twenty-one per hour while being watched.³ A separate cohort study comparing overt and covert observation found compliance of 78 percent under overt observation against 55 percent under covert observation, and, importantly, found that the size of that inflation varied by role and by setting rather than applying as a uniform correction.¹⁴

So the honest position is this. Structured observation beats self-report decisively. Unannounced measurement beats announced measurement. And anyone who quotes you an observation number without naming that bias is selling you something.

What Track A actually is

Three deliverables, all of them produced before the event rather than after it.

Roles documented, in writing, with names attached, so that question one of the debrief has an answer that does not depend on memory. Success defined in advance and agreed by the people who will be measured against it. And a performance agreement that establishes what will be observed and how, so that the eventual review is conducted against a record rather than against recollection.

Then I observe the event, and we hold the formal review within seventy-two hours of close.

Flight crews with the lowest error rates in simulator studies were distinguished not by talent but by the homogeneity of their communication patterns, which researchers interpreted as the adoption of a standard, conventionalized form. High-error crews showed no such convention.⁷ Predictability is not bureaucracy. It is what lets people improvise safely when the plan meets the day.

The debrief is where an organization learns. The planning session is where it decides whether there will be anything worth learning.