A retrospective is the only meeting most teams have where the subject is the team itself. That makes it the most valuable hour in the sprint and the easiest one to waste. When it goes badly it is rarely because the people are bad at it. It is because nobody is holding the shape of the conversation, so the conversation takes the shape of whoever talks most.

Here are the four failures, in the order they usually appear, and what actually fixes each one.

Why does one person end up talking for half the meeting?

Because open floor rewards whoever is quickest to fill silence, and that is almost never the person with the most useful thing to say. Seniority makes it worse: once the most senior person frames the problem, everyone else responds to their frame instead of offering their own.

The fix is to stop starting with discussion. Start with five minutes of silent writing — everyone puts down what they noticed, privately, before anyone speaks. Then go round in a fixed order, one item each, no responses until everyone has been. It feels stilted for the first two sessions and then it feels normal, and you will hear things you have never heard from people who have been in every retro for a year.

If you take one thing from this article, take this one. It is the single highest-leverage change available to most teams.

How do you get the quiet people to speak?

Not by asking them to. Being called on in a meeting where four people have already been fluent is a punishment, not an invitation, and the answer you get back will be agreement with whatever was said last.

Two things work. The first is the written round above, which removes the need to compete for the floor. The second is asking narrower questions. “Anything else?” produces nothing. “What was the most frustrating half hour of the last two weeks?” produces something from almost everyone, because it asks for a specific memory rather than a considered opinion.

Watch the balance rather than trusting your sense of it. Most people running a retro will swear everyone contributed; the timings say otherwise.

Why does the same issue come up every single month?

Because nobody is counting. Each retrospective starts from a blank page, so an issue raised for the fifth time arrives looking like an issue raised for the first time, and gets the same twenty minutes of sympathetic discussion it got last time.

Keep a record across sessions — a document, a page, anything — with three columns: what we decided, who committed to what, and what came up again. Open every retrospective by reading the previous session's commitments out loud. It takes two minutes and it changes the room, because it makes the gap between what the team says and what the team does visible without anyone having to accuse anybody.

A recurring theme, once it has been named as recurring, usually resolves in one of two ways: the team fixes it properly, or the team admits it is not going to and stops spending twenty minutes on it every month. Both are wins.

Why does nobody do what they said they'd do?

Because most retrospective actions are not actions. “Improve our handover process” is a wish. Nobody owns it, nothing is due, and no one will notice when it does not happen.

Cap it at two commitments per session, each with a named person and a date, and read them back at the end. Two things that happen beat eight things that don't, and a short list is much harder to quietly abandon.

Then check them next time. An uncompleted commitment is not a failure worth punishing — usually the thing was harder than it looked, or the wrong person owned it. But it has to be looked at, or the whole exercise becomes a ritual with no consequences, which is exactly how teams learn to stop taking it seriously.

What does a good retrospective actually look like?

Sixty minutes, roughly like this:

  1. Read last session's commitments out loud. Two minutes.
  2. Silent written round. Five minutes.
  3. Round the table, one item each, no responses. Fifteen minutes.
  4. Pick two things to actually discuss — by vote, not by volume. Twenty-five minutes.
  5. Two commitments, named owners, dates. Read them back. Eight minutes.

Common questions

How long should a retrospective be?

Sixty minutes for a two-week sprint, forty-five if the team is under six people. Longer sessions don't produce more — they produce the same content, more slowly, with worse attendance next time.

Should the manager be in the room?

If the manager is in the room, the retrospective is about what is safe to say in front of the manager. That can still be useful, but don't mistake it for candour. If you keep them in, they should speak last and least.

What do you do when nobody says anything?

Silence usually means the question is too broad or the room isn't safe. Write instead of speak: five minutes of silent notes, then read them out unattributed. It removes both problems at once.

How do you stop the same issue coming up every month?

Keep a written record across sessions and open each one by reading last time's commitments aloud. A recurring issue is only invisible when nobody is counting.

I'm building Facilitable, which does the above automatically for groups that meet regularly — keeps time, runs the rounds, and keeps the record so recurring themes stop being invisible. It isn't ready yet. If you'd like to see it when it is:

No product yet, no newsletter, no list swaps. One email when there’s something to try.