The Review That Changes Nothing

Teams that repeat a failure are usually not failing to reflect. They reflect constantly. What they never change is who decides the thing that went wrong.

Find the last written record of a review your unit ran after something went badly. A debrief, a post-incident summary, a lessons learned document, whatever your organization calls it. Read the list of findings at the bottom. Then ask one question about each line: did anything about who decides, or what they are allowed to decide without checking, change because of that line?

For most of these documents the answer is no, for every line. The review was real. People were candid, the sequence of events got reconstructed accurately, and somebody wrote down what should be different next time. Six months later the same conditions produced a version of the same failure, and the unit ran another review that reached largely the same findings, which is the detail worth sitting with. The second review did not fail to notice the problem. It noticed it again.

A finding is not a change

Here is what a review actually produces. It produces a shared account of what happened and a list of things people believe should be different. That list then goes out into a week that is already full, attached to nobody in particular, competing with work that has dates on it.

Consider what a line on that list is asking for. "Communicate earlier across teams" requires somebody to decide what earlier means, to have the standing to say so, and to absorb the friction the first three times they enforce it. Nothing in the review supplied any of that. The finding named a behavior and left out the authority, and behavior without authority behind it lasts about as long as the discomfort of the failure does. Roughly a month, and less if the unit is busy.

The reason the same failure returns is not that the lesson was forgotten. Everyone remembers it. The conditions that produced it are still in place: the same person still needs three approvals to move, the same decision still sits with somebody two levels from the information, the same team still has responsibility for an outcome without control over the input. A review that does not touch any of those has changed the vocabulary people use to describe the failure and nothing else.

What you will see instead of the real problem

The same finding, appearing in review after review. If you can put your hands on two years of these documents, read them in one sitting. You will usually find a small number of findings that recur with different wording, and that recurrence is the most useful thing in the pile. A lesson that has been learned three times has not been learned once. It is a structural condition being misfiled as a behavioral one.

Action items that are reminders to be careful. Watch for the ones that begin with be more, ensure that, or remain aware of. These pass through a review meeting easily because nobody can argue with them, and they are unassignable by construction. There is no version of "ensure earlier escalation" that anyone can complete, report on, or be held to. It survives the meeting precisely because it commits nobody.

A review that stops at the edge of authority. Notice where the analysis ends. It will trace the failure back through the sequence of events with real rigor and then stop just short of the decision rights that shaped it, usually because the next step would name a person in the room, or the person the room reports to. Everyone in the meeting feels the stopping point. The document records the analysis as complete.

Three ways to make a review land

Require every finding to name a decision and its owner. Not a behavior, a decision. Which call went wrong, who currently holds it, and where it should sit instead. If a finding cannot be written that way, it is an observation, which is fine, but file it separately so it stops taking up space where changes are supposed to be. A review that produces two findings with owners is worth more than one that produces eleven without them.

Put a date on it and check the date. Every change gets a person and a day, and the day gets checked whether or not anything went wrong in the meantime. The check is the whole mechanism, because the reason lists decay is that nothing ever asks about them again. Ten minutes in an existing meeting is enough. What you are looking for is whether the change actually got made, not whether the person feels behind on it.

Read the last review before you run the next one. Open the previous document at the start of the meeting and go through its findings out loud. If a finding is repeating, say so plainly and stop discussing the incident, because the incident is no longer the subject. The subject is why a change your unit agreed to did not happen, and that is a different conversation with a different answer, usually about capacity, authority, or somebody's unwillingness to carry the friction.

The part of this that is yours

Some findings do not get written down because writing them down means naming your own decision. The approval you kept, the exception you granted in March, the reorganization you did not want to revisit, the person you left in a role that was structured to fail. Your team knows which finding is missing. They watched the meeting stop at the edge of it.

You do not have to hold a public accounting for that. What you do have to do is stop treating a review as a thing your team runs and you receive. If the only findings that survive are the ones that require nothing from you, the unit learns something more durable than any lesson on the list, which is that the review is a procedure rather than a way anything changes.

Three questions

Take your last three reviews. How many findings from the first two were actually completed, and can you name who completed them without asking anyone?

Which finding has appeared in more than one review? Write it down, and then write down the decision right that would have to move for it to stop appearing.

At the last review you sat in, where did the analysis stop? Ask yourself what the next question would have been, and who in the room would have had to answer it.