Delivery field notes

How do you run a lessons learned meeting that works?

Run a lessons learned meeting that changes the next project: a timeline before opinions, causes before blame, and each lesson turned into an owned change.

Why most lessons learned are never learned

The project ends, someone books a lessons learned meeting, and the team fills a whiteboard with sticky notes: "communication could be better", "start testing earlier", "supplier was late". Someone types them up. The document goes into a shared folder. Six months later the next project makes the same mistakes, and nobody opens the folder.

The meeting was not the problem. The output was. A list of observations is not a lesson. A lesson is a change: to a template, a checklist, a contract clause, a planning rule, a training session. If nothing changes, nothing was learned.

The UK government's Orange Book, HM Treasury's guidance on managing risk, puts the purpose plainly: all strategies, policies, programmes and projects should be subject to proportionate evaluation where practicable, because learning from experience helps avoid repeating the same mistakes and spreads improved practice. NASA's public Lessons Learned System shows what the output looks like when it works: each lesson describes the original driving event and gives recommendations that feed into training, best practices, policies and procedures.

Before the meeting: build the timeline

Opinions about what went wrong are shaped by the last two weeks of the project, which are usually the most stressful. Start from facts instead. Before the meeting, the project lead builds a one-page timeline from the plan, the status reports, the change log and the risk register:

  • Planned and actual dates for each milestone.
  • When each major risk was first raised, and when it became an issue.
  • Each approved change, with its date and effect.
  • Decision points: who decided what, when.

Send it to attendees a day before and ask each person to mark one moment where things went better than expected and one where they went worse. This takes ten minutes each and gives the meeting a shared starting point.

In the meeting: four questions, in order

A ninety-minute meeting for a team of six to eight people is enough for most projects. Keep the order. It stops the meeting from jumping straight to blame or solutions.

  1. What did we plan, and what happened? Walk the timeline. Stick to facts. This takes about twenty minutes.
  2. Where did the outcome differ from the plan, and why? For each significant gap, ask "why" until you reach something the organisation controls: a process, a template, a contract term, a planning assumption. "The supplier was late" is a fact. "Our contracts had no delivery-date confirmation step" is a cause you can change.
  3. What went better than expected, and why? Good surprises are lessons too, and they are the ones most often lost. If one team member's habit saved the schedule, it should become the default.
  4. What will we change, who owns it, and by when? Each lesson ends here or it does not leave the room.

Keep people out of causes. "Sam didn't chase the supplier" is a person. "Nobody owned supplier confirmation, so chasing depended on whoever noticed" is a system. Talking about systems keeps people candid, and candour is the only thing that makes the meeting worth holding.

A worked example

The project below is fictional: an office move that went live two weeks late. Here is one lesson, taken from observation to change.

StepContent
ObservationCard readers arrived 13 days after the date in the supplier's quote. Go-live moved from 2 Dec to 16 Dec.
Timeline factsRisk raised 21 Oct ("supplier may be late"). Response column read "Monitor" for five weekly reviews. Delivery date never confirmed in writing. Fallback (rental readers) quoted 12 Nov, too late to order.
CauseThe risk had an owner but no trigger date, so nobody knew when the fallback had to be ordered. Supplier quotes gave lead times, not confirmed dates.
What went wellFacilities found a compatible rental supplier in one day once asked.
Change 1Risk register template gets a required "trigger date" field for any risk with a fallback. Owner: PMO lead. Due: 31 Jan.
Change 2Purchase orders for critical-path equipment require a written delivery date, confirmed within five working days. Owner: procurement. Due: 28 Feb.
Change 3Add the rental supplier to the approved list, so the fallback can be ordered without a new supplier check. Owner: facilities. Due: 31 Jan.
CheckAt the next project's kick-off, confirm the register has trigger dates and POs have confirmed delivery. Owner: next project's lead.

Notice the last row. A lesson is not finished until someone checks that the change reached the next project. Without that check, changes 1 to 3 are just better-worded sticky notes.

After the meeting: put lessons where people will find them

A lessons learned document in a folder is read by almost nobody. Put each change where the next project will meet it:

  • In templates: a new field in the risk register or status report.
  • In checklists: the kick-off checklist, the procurement checklist.
  • In contracts and purchase orders: a clause, a confirmation step.
  • In the project start-up: a fifteen-minute review of the last three projects' changes at each kick-off.

Then keep a short register of the changes themselves, with owner, due date and status, and review it monthly until each one is done. NASA's approach, where lessons feed training, practices and procedures, works because the lesson ends up in the thing people already use.

Facilitating the meeting

The facilitator's job is to keep the meeting on facts and causes, and to make sure every lesson ends with an owner. A few habits help. Read the timeline aloud at the start so everyone hears the same version. When someone names a person, rephrase it as a system question: "What made that step depend on one person noticing?" Cap discussion of any single gap at fifteen minutes; if it needs more, it needs its own follow-up. Close the meeting by reading back each change, its owner and its date, and ask the owners to confirm aloud.

Common objections

"We're too busy to hold a meeting at the end." Hold a shorter one, forty-five minutes, with the timeline and question 4 only. A smaller output that changes something beats a thorough review that changes nothing.

"People won't be honest." They will be more honest if the questions are about systems and the facilitator is not the person who made the key decisions. Consider asking someone from outside the team to run it.

"We already do retrospectives every sprint." Sprint retrospectives improve how the team works next sprint. A project lessons learned review looks at things that outlive the team: contracts, templates, procurement rules and governance.

Try it on a project you know

Take a project you were part of, finished or nearly finished, and spend fifteen minutes:

  1. Write a five-line timeline: planned against actual dates for the main milestones.
  2. Pick the largest gap and ask "why" until you reach something the organisation controls.
  3. Write one change, with an owner and a due date.
  4. Name where the change will live so the next project meets it.

The recovery plan example shows the decisions you will want on the timeline, and the RAID log example shows where risks, assumptions, issues and dependencies should have been recorded during the project.

To practise the risk habits that most lessons learned reviews ask for, try the open exercise Municipal permit portal: risk register.

Sources

Method background

These sources provide general background. The worked examples and worksheets are original Response Shift teaching material; the organisations do not endorse this product.