- Description
- Une galaxie spirale barrée photographiée par le télescope spatial Hubble, ses bras enroulés autour d'un noyau lumineux.
- Source
- NASA, ESA et l'équipe Hubble Heritage (STScI/AURA) · original page
- Licence
- PD Public domain
Continuous Improvement and Learning (AMC)
This is the multiplier theme: without it, an organization repeats its mistakes; with it, each project improves the next ones. It is also the one most often simulated: retrospectives where…
Level: 🟢→🔴 · Code: AMC · Matrix: PI
Why this theme#
This is the multiplier theme: without it, an organization repeats its mistakes; with it, each project improves the next ones. It is also the one most often simulated: retrospectives where people talk without acting, post-mortems that point to someone to blame, “lessons learned” archived and never revisited. Maturity is visible in one thing: concrete, tracked changes whose impact is verified.
What this theme measures#
- Axis: from forgetting to organizational learning that simplifies as much as it adds.
- Key indicator: share of improvement actions closed on time and with confirmed impact; recurrence rate of incidents with the same cause.
- Unit: % of actions closed; % of recurring incidents; number of rules removed or simplified per year.
The five levels#
| Level | Short name | Promise | Typical indicator (orders of magnitude) |
|---|---|---|---|
| 1 | No feedback | The same problems keep coming back; no one has time to stop. | Recurrence > 50% |
| 2 | Retrospectives | People occasionally meet to discuss what was difficult; a few ideas come out of it. | One meeting per quarter; actions closed < 30% |
| 3 | Tracked actions | Retrospectives and incidents produce actions with an owner and due date, reviewed in the next session; no blame-seeking. | Actions closed 60 to 80%; recurrence 20 to 40% |
| 4 | Measured impact | The impact of improvements is verified with indicators; root causes are addressed; lessons are shared across teams. | Actions closed > 80%; impact confirmed on ≥ 60%; recurrence < 20% |
| 5 | Learning organization | The organization experiments, removes what is no longer useful, shares what it learns, and its rules evolve on their own. | Recurrence < 10%; rules removed each year |
Milestones by level#
| ID | Milestone | Verifiable evidence |
|---|---|---|
| PI-AMC-1.1 | Major incidents are shared with the team. | Record of a post-incident feedback session. |
| PI-AMC-1.2 | A completed project is followed by a closeout (even brief). | Closeout document. |
| PI-AMC-1.3 | Lessons are documented somewhere. | Lessons page or file. |
| PI-AMC-2.1 | A retrospective is held on a fixed cadence (per iteration or per quarter). | Calendar and meeting notes. |
| PI-AMC-2.2 | Incident post-mortems use a common format. | Template used on ≥ 3 incidents. |
| PI-AMC-2.3 | Actions are logged with an owner. | “Owner” column filled in. |
| PI-AMC-3.1 | Each action has a due date and is reviewed in the next session. | Action list with recorded review. |
| PI-AMC-3.2 | Post-mortems are blameless and focused on conditions. | Review of 3 post-mortems: no person named as the cause. |
| PI-AMC-3.3 | Action closure rate is tracked (≥ 60%). | Closure indicator. |
| PI-AMC-3.4 | Time is reserved for improvement (for example 10% of capacity). | Reserved capacity visible in the plan. |
| PI-AMC-3.5 | Lessons are accessible to the team (knowledge base, single page). | Up-to-date page, consulted. |
| PI-AMC-4.1 | Each significant improvement defines an expected-impact indicator before launch. | Action sheet with target indicator. |
| PI-AMC-4.2 | Impact is verified after a defined period and documented (effective, partial, none). | Impact review on ≥ 5 actions. |
| PI-AMC-4.3 | Root cause analyses (5 whys, cause tree) are conducted on major incidents. | Analysis sheets. |
| PI-AMC-4.4 | Lessons are shared across teams (cross-review, community of practice). | Cross-team sessions; adoption of a neighboring practice. |
| PI-AMC-4.5 | Recurrence of incidents with the same cause is measured. | 12-month recurrence rate. |
| PI-AMC-5.1 | Structured experiments (hypothesis, duration, criterion) are run regularly. | Experiment log. |
| PI-AMC-5.2 | Obsolete rules, steps, or reports are removed each year. | List of justified removals. |
| PI-AMC-5.3 | Improvement indicators are reviewed to prevent gaming. | Annual indicator review with replacements. |
| PI-AMC-5.4 | The year's projects feed a shared retrospective across the whole organization. | Annual presentation and distributed materials. |
Detailed reading by level#
Level 1. Signs: fix it, move on. Gain: maximum production time in the short term. Does not guarantee sustainability. Difficulty: 1/5. Effort: 1/5.
Level 2. Signs: retrospectives are pressure-release valves; people say “we should have.” Gain: awareness. Does not guarantee change. Difficulty: 2/5. Effort: 1/5.
Level 3. Signs: each session starts with actions from the previous one; the team dares to talk about mistakes. Gain: fewer recurring problems. Does not guarantee actions are effective. Difficulty: 3/5. Effort: 2/5.
Level 4. Signs: an action without an impact indicator is suspicious; you know what worked. Gain: the organization invests in what works. Does not guarantee innovation. Difficulty: 4/5. Effort: 3/5.
Level 5. Signs: experiments, removals, sharing. Gain: fast adaptation to changing context. Does not guarantee absence of cultural drift. Difficulty: 5/5. Effort: 4/5.
Partial achievement and rating#
Common rule: see README.
- 0.5: regular retrospective but actions not tracked; post-mortems done without a template.
- 1: three consecutive tracking cycles are available for review.
- A partial 5: verified impact of improvements, cross-team sharing, one rule removed per year; no experiment log. Score ≈ 4.6.
Incident: a release package was delivered without the migration script for a parameter. Blameless post-mortem: the release checklist did not include this item; action: add it and verify by script. Impact indicator: zero deliveries with missing items over 6 months. Result reviewed in session: achieved. This type of loop closes the circle (see Delivery).
Retrospective theater. Symptom: the same points come back session after session. Cause: no action is owned or arbitrated by governance. Remedy: limit to 1–3 actions, assign an owner, report back in the next session.
Goodhart's law: using action closure rate as a target produces trivial actions that are easy to close. Complement with verified impact (PI-AMC-4.2).
Also count removals: an organization that only adds rules gets heavier; one that removes as much as it adds stays agile. Direct link with adaptive governance (02).
Moving to the next level#
Possible means (optional)
- 1 → 2: one retrospective per quarter; one post-mortem template.
- 2 → 3: actions with due date and review; blameless post-mortems; reserved time; see Scrum in Practice and Agile and Lean Methods (Manifesto, Kanban, XP, Lean, Shape Up, DSDM, SAFe, DevOps…).
- 3 → 4: impact indicators, root cause analysis, community of practice; Project Metrics and Reporting.
- 4 → 5: experiment log, yearly removals, cross-project feedback.
Can group retrospective points, identify recurrences in post-mortem history, suggest causal hypotheses; it does not replace psychological safety, without which facts are not spoken. See AI for Project Management (General, Then Data).
See also#
- Related themes: 06 Team · 05 Risks · 02 Governance · 11 Budget and Value.
- Other matrices: Technical Debt and Maintainability (DL) · Observability and Reliability (DL) · Operations and Support (data).
- Practice: Scrum Framework (Scrum Guide 2020) · Agile and Lean Methods (Manifesto, Kanban, XP, Lean, Shape Up, DSDM, SAFe, DevOps…) · Project Management Pitfalls and Best Practices (Cross-Cutting Summary) · Glossary.
✻ Written by Claude Sonnet 5.5
