docs: make M01 cohort executable

docs/validation/M01-COHORT-RUNBOOK.md

@@ -0,0 +1,340 @@
+# M01 Cohort Runbook — first 5 learner sessions
+
+## Purpose
+
+Run the first five real M01 learner-validation sessions consistently enough to make one product decision: **Promote / Revise / Reject**.
+
+This is an operational product gate for a small exploratory cohort, not a statistical effectiveness claim.
+
+Published learner experience under test:
+
+- course: `https://tim8es.github.io/pmo01/`
+- validation route: `https://tim8es.github.io/pmo01/#/validation/m01`
+
+Do not activate M02 during this cohort.
+
+## Cohort plan
+
+Run the same core sequence for participant IDs `P01` … `P05`.
+
+Eligible participant: someone who regularly reasons about delivery/project work, such as PM, Delivery/Process Manager, Team Lead with delivery ownership, or comparable product/operations role.
+
+Do not count the course author or synthetic/AI sessions.
+
+Plan roughly 100–130 minutes per completed session. Time is an operating estimate, not a validity criterion. A session is valid only if the required evidence below is complete.
+
+| Session | Participant | Status | Valid / excluded | Reason if excluded |
+|---|---|---|---|---|
+| P01 | | planned | | |
+| P02 | | planned | | |
+| P03 | | planned | | |
+| P04 | | planned | | |
+| P05 | | planned | | |
+
+If a session is excluded, document why and run an additional participant so the final review still contains at least five valid sessions.
+
+---
+
+## Exact session script
+
+### 0. Setup — 2 minutes
+
+For **each new participant**, use a fresh browser context where both PMO01 localStorage keys are absent. Use a fresh browser profile/private session or clear all PMO01 site data before opening the validation route.
+
+Do **not** prepare a different participant by pressing only **Сбросить M01 validation data**. That button clears the validation answers but deliberately preserves the main-course progress, so reused M01 progress could contaminate the next participant.
+
+1. Start a fresh browser context or clear all PMO01 site data.
+2. Open the validation route.
+3. Confirm that no previous baseline result or completed M01 state is visible.
+4. Tell the participant:
+
+> Проходи задания так, как решал бы реальную рабочую ситуацию. Я не буду подсказывать правильный ответ. Если формулировка или интерфейс мешают — говори вслух, что именно непонятно.
+
+Do not explain the seven-flow model or diagnostic technique before baseline.
+
+### 1. Baseline — 10–15 minutes
+
+Participant reads **«Релиз снова сдвинулся»** and, without coaching:
+
+1. writes the free-text diagnosis;
+2. answers all five rubric questions;
+3. submits baseline.
+
+Baseline dimensions, each `0..3`:
+
+- mechanism;
+- evidence;
+- trade-offs;
+- intervention;
+- change condition.
+
+Required evidence:
+
+- baseline total `0..15`;
+- five dimension scores;
+- one-sentence summary of the participant's diagnosis;
+- whether they mainly reasoned from people/symptoms or from a repeatable mechanism;
+- any wording/UI ambiguity.
+
+**Blinding rule:** do not reveal baseline score or option feedback before post-case submission.
+
+### 2. M01 learning experience — about 80 minutes
+
+Participant completes, in order:
+
+1. `project-system` Learning Lab;
+2. its **cold decision drill** — first choice is frozen when feedback opens;
+3. its exit drill;
+4. workbook/transfer fields required by the lesson;
+5. `system-diagnostic` Learning Lab;
+6. its **cold decision drill** — first choice is frozen when feedback opens;
+7. its exit drill;
+8. workbook/transfer fields required by the lesson.
+
+The validation route does **not** ask the learner to repeat those cold drills. After both lessons satisfy their evidence-aware completion gates, return to validation and continue directly to post-case.
+
+Facilitator does not teach the answer. Record only observed friction and reasoning changes.
+
+For each cold decision drill capture:
+
+- first choice;
+- short reason for the choice;
+- whether feedback changed the reasoning: `yes / no / unclear`;
+- whether the drill exposed a misconception: `yes / no / unclear`;
+- usefulness: `useful / partial / weak`.
+
+### 3. Post-case — 10–15 minutes
+
+Participant completes **«Интеграция к запуску партнера»** without baseline feedback having been shown beforehand.
+
+Required evidence:
+
+- post total `0..15`;
+- five dimension scores;
+- total delta `post - baseline`;
+- number of improved dimensions;
+- free-text diagnosis summary;
+- whether free-text reasoning supports the scored result.
+
+Individual positive prototype signal is unchanged:
+
+```text
+post >= baseline + 3
+AND
+at least 2 rubric dimensions improved
+```
+
+This is a development signal, not mastery.
+
+### 4. Real-project transfer — 10–15 minutes
+
+Participant uses a current or recent real project and records:
+
+1. visible symptom;
+2. repeatable mechanism/system condition;
+3. one minimal intervention;
+4. one early observable signal;
+5. observed evidence, or explicitly `not tested yet`;
+6. next decision that follows from the evidence.
+
+Classify transfer:
+
+- **credible** — mechanism is not merely a person/symptom; intervention acts on that mechanism; early signal is observable; next decision follows from evidence;
+- **partial** — chain is directionally useful but one important link is weak/missing;
+- **restatement only** — course language is repeated without a defensible real-project causal chain;
+- **not completed**.
+
+Do not require confidential names, customer data, credentials, commercial terms, or sensitive metrics.
+
+### 5. Reflection + interview — 8–10 minutes
+
+Ask exactly these five questions:
+
+1. Что ты сначала считал главной проблемой?
+2. Что изменилось в диагнозе после M01?
+3. Какая часть M01 реально изменила ход мысли?
+4. Где формулировка, feedback или интерфейс мешали или казались неубедительными?
+5. Где ты сможешь применить это завтра без открытия курса?
+
+Capture paraphrased answers. A verbatim quote is optional.
+
+### 6. Facilitator closeout — 3 minutes
+
+Mark:
+
+- response/state loss: `yes / no`;
+- route/navigation confusion: `yes / no`;
+- blocking readability/interaction issue: `yes / no`;
+- rubric ambiguity that could change score: `yes / no`;
+- session: `valid / exclude`.
+
+Exclude only when evidence is materially contaminated, for example:
+
+- baseline was not blind/fresh;
+- required baseline or post-case evidence is missing;
+- facilitator coached the answer materially;
+- a reliability/UI failure prevented completion or corrupted answers.
+
+Ordinary disagreement with the course, low score, or weak transfer is **not** a reason to exclude.
+
+---
+
+## Minimal evidence form — copy once per participant
+
+```text
+Participant ID: P0_
+Role / experience:
+Date:
+Valid / excluded:
+Exclusion reason (if any):
+
+BASELINE
+Total /15:
+Mechanism /3:
+Evidence /3:
+Trade-offs /3:
+Intervention /3:
+Change condition /3:
+Diagnosis summary:
+Observed pattern: people/symptom / mixed / mechanism
+Ambiguity/friction:
+
+DRILL 1 — project-system cold drill
+First choice:
+Reason:
+Feedback changed reasoning: yes / no / unclear
+Misconception exposed: yes / no / unclear
+Useful: useful / partial / weak
+
+DRILL 2 — system-diagnostic cold drill
+First choice:
+Reason:
+Feedback changed reasoning: yes / no / unclear
+Misconception exposed: yes / no / unclear
+Useful: useful / partial / weak
+
+POST-CASE
+Total /15:
+Mechanism /3:
+Evidence /3:
+Trade-offs /3:
+Intervention /3:
+Change condition /3:
+Delta:
+Improved dimensions count:
+Individual signal: positive / absent
+Diagnosis summary:
+Free text supports score: yes / partial / no
+
+TRANSFER
+Mechanism credible: yes / partial / no
+Intervention acts on mechanism: yes / partial / no
+Early signal observable: yes / partial / no
+Next decision follows from evidence: yes / partial / no
+Overall transfer: credible / partial / restatement only / not completed
+
+UX / RELIABILITY
+State loss: yes / no
+Route/navigation confusion: yes / no
+Blocking readability/interaction issue: yes / no
+Rubric ambiguity could affect score: yes / no
+Main friction:
+
+INTERVIEW
+What changed in diagnosis:
+What caused the change:
+What felt weak/misleading:
+Where they can apply it tomorrow:
+
+SESSION SUMMARY
+Learning signal: positive / mixed / absent / uninterpretable
+Transfer: credible / partial / absent
+Interaction usefulness: strong / mixed / weak
+Major friction: none / manageable / blocking
+Recommended action: keep / revise / investigate
+```
+
+The longer `M01-SESSION-RECORD-TEMPLATE.md` remains available when a session needs richer notes. This minimal form is the required first-cohort evidence set.
+
+---
+
+## Cohort decision after 5 valid sessions
+
+First calculate:
+
+- number of positive individual signals;
+- number of credible transfers;
+- for each cold drill, number of sessions where it was `useful` or `partial` and why;
+- number of sessions with blocking UX/reliability problems;
+- number of sessions where rubric ambiguity could plausibly affect the result;
+- repeated content/wording problems appearing in at least two valid sessions.
+
+### PROMOTE to Phase 2
+
+Choose **Promote** only if all are true:
+
+1. at least 5 valid completed sessions;
+2. at least **3 of 5** show the positive individual signal (`+3` and ≥2 improved dimensions);
+3. at least **3 of 5** produce **credible** real-project transfer;
+4. qualitative free-text/interview evidence does not materially contradict the score improvement;
+5. no unresolved blocking reliability/UX defect materially distorted more than one valid session;
+6. no repeated rubric ambiguity plausibly explains the observed gains;
+7. no unresolved learner-driven content-model exception blocks representing the validated M01 contract.
+
+A drill may be revised later; one weak drill alone does not block Promote if the module-level reasoning and transfer gates pass and the weak interaction is clearly identified.
+
+### REVISE and retest
+
+Choose **Revise** when the mechanism remains plausible but any of these is true:
+
+- only **2 of 5** show the positive individual signal;
+- transfer is mostly partial rather than credible;
+- the same wording/content/interaction problem appears in at least **2 valid sessions** and could affect reasoning or measurement;
+- a rubric item creates plausible false-positive/false-negative results in at least **2 valid sessions**;
+- reliability/usability interferes with otherwise interpretable learning evidence;
+- qualitative reasoning improves but the current scored instrument fails to capture it consistently.
+
+Record the smallest specific revision and run a new cohort or targeted retest before Phase 2.
+
+### REJECT mechanism
+
+Choose **Reject** only when the sessions are valid enough to interpret and both are true:
+
+1. **0–1 of 5** show the positive individual signal; and
+2. **0–1 of 5** show credible real-project transfer;
+
+and there is no blocking measurement/usability defect that could reasonably explain the failure.
+
+Also consider Reject for a specific interaction when it repeatedly adds effort without exposing misconceptions, changing reasoning, or contributing to transfer. Rejecting one interaction does not automatically mean rejecting the whole M01 module.
+
+### Borderline/conflicting cohort
+
+If evidence does not fit the rules cleanly — for example 3/5 score signals but only 1/5 credible transfer — choose **Revise and retest**, not Promote.
+
+When quantitative and qualitative evidence conflict materially, choose **Revise and retest** and document the conflict.
+
+---
+
+## Five-session cohort summary
+
+| ID | Baseline | Post | Delta | Improved dims | Positive signal | Transfer | Drill 1 | Drill 2 | Blocking friction |
+|---|---:|---:|---:|---:|---|---|---|---|---|
+| P01 | | | | | | | | | |
+| P02 | | | | | | | | | |
+| P03 | | | | | | | | | |
+| P04 | | | | | | | | | |
+| P05 | | | | | | | | | |
+
+Final decision:
+
+```text
+PROMOTE / REVISE / REJECT
+
+Why:
+
+Evidence:
+
+Required next action:
+```
+
+Then complete the existing `M01-COHORT-REVIEW-TEMPLATE.md` for the durable Phase 1 decision record and update Issue #4.