It is easy to explain a failed project after the outcome is known. It is harder, and more useful, to reconstruct what the team understood when it made each decision. I try to begin there rather than collecting a list of mistakes that now look obvious.
Separate the decision from the outcome
A reasonable decision can have a disappointing result, especially when important information arrives late. I ask which signals were available, how concerns were discussed, and what made a change of direction difficult. That produces lessons about the working environment rather than a judgment about individual competence.
A good retrospective makes the next difficult conversation easier.
The most useful outcome is usually a small change in practice. Earlier feedback from operations or a clearer stopping condition may matter more than a new process document. I want the lesson to survive the meeting and show up in the next project when it can still help.