How to run a team feedback debrief in two hours

Feedback does not always help. In the largest review of feedback interventions, Kluger and DeNisi (1996) found that feedback made performance worse in more than a third of measured cases (38%, across 12,652 participants). Feedback that draws attention to the self rather than the task is the kind most likely to backfire.

Debriefs have a stronger record. Tannenbaum and Cerasoli (2013) pooled 46 samples (2,136 participants) of teams and individuals who debriefed against those who did not, and found an effect of d = 0.67. In the studies reviewed, the chance that a team that debriefed outperformed a comparable team that did not was about 2 in 3. The authors describe the effect as a 20–25% improvement; that figure is a percentile shift, not a gain in output, which matters if you repeat it to your own team.

This is the agenda the product generates from a team’s own results: about two hours, in four phases — review the results, discuss, make a plan, schedule the follow-up — with a break between the review and the rest. The minute counts are fixed; the content of each block is filled from the data.

Open with three sentences, and say them even if everyone has read the pack. These are the team’s own ratings of the team, averaged. Individual ratings are not shown. The results are based on however many of you submitted — if that was nine of twelve, say nine of twelve, and say what the completion means for how much weight the numbers can carry. Then the sentence that does the most work in the meeting: where members disagree, the pack says so, and those scales are a conversation, not a verdict. Close the block by saying how the time will run: the results first, a break, then what the team does about them, with the follow-up booked before anyone leaves.

Start with strengths and spend the whole block on them. A team that hears its strengths named specifically — “Coordination and process: hand-offs go smoothly, and people agree that they do” — has something to protect when the harder blocks arrive, and the facilitator has a reference point for later: “we said we were good at this; is that still true in the situation we are describing?”

The pack names up to three scales with the most room. For each, put the spread on the table before the mean. A scale with a low mean and tight agreement points to a shared experience. A wider spread means people gave different ratings, so the useful question is what might explain the differences? rather than why is this low? Resist diagnosis. The questionnaire measures behavior and climate; it does not know why, and neither does the pack.

Besides rating the team, each person answers a few questions about how things are for them: whether their workload is sustainable, whether they feel safe to speak up, whether the team’s goal is worth their best effort, whether they are learning, belong and make progress on work that matters, and whether they expect to still be on the team in a year. These are not ratings of the team, so there is no agreement figure. The pack gives each a mean and a count — “Sustainable workload: 3 of 7 answered below the midpoint” — and raises one here when two or more people answered on the disagree side, or when its mean is under the midpoint. Read the count before the mean. Three people saying their workload is not sustainable is a finding even when the average looks fine.

The question for the room is what would have to change for that to move?, never who are the three? Nobody’s answers are shown, and in Facets a self-report is not reported at all until 4 people have answered it.

The network picture shows who people say they go to for help and for a second opinion. Read it as load and access, not popularity. A person everyone goes to is carrying something; a person nobody goes to may be new, remote, or in a role that does not generate requests. The question for the room is whether the pattern matches how the team wants work to flow, and what one change would move it.

Take the break, and take all of it. The first half of the meeting was about what the results say; the second half is about what the team will do, and it asks different work of people. A break also gives anyone who found a block hard a few minutes before they are asked to speak again.

One rule for it: the results are not discussed in twos and threes in the corridor. A side conversation about who might have answered what is the performance conversation the meeting was built to avoid, held where nobody can stop it. Whatever someone wants to say about the results, they say back in the room.

Start from the shared team report. Every private note ends with two lines written for this block: one thing to share, and one question to ask teammates. People may choose to share a takeaway from their own note, so go round once. Each person says one thing they will keep doing and one they would like to do more of, in their own words and only if they choose to; passing is a full answer. Nobody is asked to reveal ratings or to work out who said what, and the facilitator stops any attempt to work back from what someone shares to who rated them how.

Then turn to the pack’s questions for the conversation. They are written from this team’s own pattern — a scale members are split on, a self-report several people answered low — and they are where the room’s own explanations belong, before anything is decided.

The pack proposes two commitments to start from. Each is a sentence with an indicator: “We will start each planning meeting with a two-minute risk round. Indicator: risks raised before the deadline, not after.” Two is a deliberate starting point, because a short list is more likely to be done than a backlog. The team can add its own commitments, and each one needs the same treatment. Write each down as an if-then plan with an owner and a check date: if it is a planning meeting, then it opens with a two-minute risk round; one named person checks the indicator, on a date six weeks out. The next block puts that date in the calendar.

A commitment without a date is easy to forget. Before anyone leaves, record the commitments where the team can find them — in Facets, on the run’s page, each with who checks it and when. Then put the six-week check in the calendar while everyone is in the room. About six weeks after release, a team lead can open a short pulse: two minutes per person on how far each commitment has happened.

Last, agree when to run the whole survey again: not at the six-week check, but when the team has changed enough that the questions would get different answers. The block is short because the work was done in the one before it; if it runs long, the decisions were not really made.

Meet briefly about six weeks later, on the date the meeting’s last block put in the calendar, and ask three questions: did we do what we committed to; did the indicator move; do we keep, change, or drop each. Do not re-run the survey to answer it. Individual change is hard to measure at team size, and a second run six weeks on would mostly measure noise. What Facets offers instead is a two-minute pulse that a team lead can open: everyone says how far each recorded commitment has happened, and answers the same two questions about speaking up and workload the run asked, so the check-in starts from the team’s own answers rather than impressions. Run the survey again when the team has changed enough that the questions would get different answers — a new quarter, a new composition, a new mandate.

The private notes each person received are the other half of the loop. A note is for the person it is about, and a facilitator the team has named can also read it (see who can see what). Individual peer scores appear only from 3 raters. What a person does with their note is theirs to decide; the self–peer gap is a good place to start.