Held at closure, a project lessons-learned review comes at the moment the team disperses. Those who have learned something move on to other work, the next project may start with other people, and the report risks ending up in a file nobody has any reason to reopen. The exercise can therefore be well run and change nothing.
We suggest holding it when three conditions are met: it takes place at a time when its conclusions can still be of use, it focuses on a bounded subject, and a designated person has the power to act on them. This guide takes them in order (timing, subject and participants, follow-up), then describes the cases where it is better to forgo it than to rush it.
When a project lessons-learned review is still useful
The term, abbreviated in French as REX, covers several practices, which differ first of all in their timing.
In the Scrum framework, the exercise is built into the team’s rhythm. The Scrum Guide (scrumguides.org, November 2020 version, read on 12 November 2026) calls the event that concludes each Sprint the Sprint Retrospective. Its purpose is “to plan ways to increase quality and effectiveness”. The Scrum Team inspects how the last Sprint went “with regards to individuals, interactions, processes, tools, and their Definition of Done”, then identifies the most helpful changes. The Guide adds that the most impactful improvements “are addressed as soon as possible” and may be added to the Sprint Backlog for the next Sprint. Its duration is capped: three hours at most for a one-month Sprint.
Two other forms belong to common usage, and we present them as such. The end-of-project review, or end-of-phase review, brings the team together once the work has been delivered. The post-incident review takes place in the days following an outage or a failed cutover, without looking for someone to blame; our page SRE and observability describes it for operations.
These forms differ in the distance between the exercise and the next opportunity to apply what it teaches. The retrospective finds this opportunity in the next Sprint, with the same team. The closure review only finds it in the next project, if it resembles the first. The decision criterion follows from this: place the exercise where follow-up remains possible. On a long project, this means holding it at the end of a phase, after a cutover or after a missed milestone, without waiting for closure. The facts are recent, and the next phase can benefit from them.
A bounded subject, and the people to bring together
A lessons-learned review that tries to cover everything risks producing a list of regrets. We suggest bounding the subject to a few questions chosen before the session. For example: why the estimate for data migration turned out to be too short; what enabled the cutover to go as planned; at what point the dependency on the neighbouring system could have been seen.
The useful material is the gap between what the plan assumed and what happened, with the information available to those who decided, at the moment they decided. What worked is examined on the same footing as what went wrong.
The session brings together those who did the work and those who decided: project team, user representatives, operations, the supplier where the relationship allows. The candour of what is said depends on the facilitation and on the rule announced. Facilitation falls to a person outside the project’s line of management, for example a member of the project office, or PMO, or a project manager from another team. The rule is stated when opening the session: it concerns situations and decisions, and leaves aside the assessment of people.
Neighbouring exercises have a different purpose. The project review decides the next steps of a project in progress, before those with authority over it. The project audit establishes the facts: a third party independent of the project checks them against the evidence, cross-checks them and draws options from them for a sponsor. The lessons-learned review, for its part, gathers what participants take away, without verifying what they say. On a project in difficulty whose real state is disputed, asking it to do the work of an audit exposes the team to a session it may take for a trial.
Few conclusions, each with a recipient
The session minutes are only a record. The exercise produces a short list of conclusions, which common usage also calls “lessons learned” and which we suggest sorting into three kinds:
- a common rule to change (an estimation model, the criteria for a milestone, a standard clause, a pre-cutover checklist), addressed to the person who owns that rule;
- a specific project to alert, the one about to encounter the same situation: the conclusion goes to its project manager, in person, at the framing stage;
- a finding with no follow-up, recorded as such so as not to suggest that an action will follow.
Each conclusion of the first two kinds carries a name, an action and a date. The name is that of a person with the power to change something. The team that is breaking up cannot play this role.
What remains is to check that the follow-up has happened. In an organisation with a project office, this responsibility may fall to it: it keeps the list of open conclusions, presents them at the framing of comparable projects and checks that the common rules have been changed. Our page on setting up a PMO describes the mandate that makes this role possible. Failing that, the sponsor designates this person before the session.
The proof that a lessons-learned review has been useful is found in the next project: an estimate made differently, a milestone whose criteria have changed, a check added before a cutover.
The cases where it is better to forgo it
A rushed lessons-learned review has a cost: it can wear down the trust of those who were asked to speak, to the detriment of the next session. Several situations justify abstaining, or waiting.
The first is the absence of a recipient. If nobody can change a rule or alert another project, the session will produce only a document.
An ongoing dispute is another. When a dispute sets the organisation against a supplier, or an internal inquiry into responsibility is under way, everyone weighs their words. The exercise waits for the outcome of the dispute; in the meantime, the facts are established by other means, with your lawyers if the contract is at issue.
The same applies when the sponsor expects a name. A session commissioned to designate a culprit changes in nature: everyone may prepare their defence there.
The moment may also have passed: the team dispersed long ago, and memories may have rearranged themselves around the final result. A few targeted interviews, on a specific question, are then better than a plenary session.
Finally, the conclusions of the previous lessons-learned review may still be pending. Holding a new session before having dealt with or closed them only lengthens the list.
Forgoing it is stated: the decision is recorded at closure, with its reason. And before putting a session in the diary, one check comes before all others, that of the recipient: their name, and what they have the power to change. As long as that name is missing, the session can wait.