Construct a retrospective flow experiment and inspect censored outcomes

A hands-on PSM-I lab. You produce the real artefact and 10 automated checks verify it behaves the way the exam expects.

Certification: PSM-I
Format: Structured configuration
Difficulty: hard
Estimated time: 40 min
Automated checks: 10

The brief

Repair retrospective-flow.json. policy:{intervention,sourceRef,hypothesis,wipLimit,targetReduction} selects the exact protect-verification/Q0 or pair-at-build/R0 model, WIP1–3 and a retained lead-reduction target>=0.10. baseline:{completed,returns,meanLeadTime} preserves the historical six-item cohort, two returns and6.5 customer lead mean. trace:[{day,admitted:[IDs],build:[IDs],verify:[IDs]}] records every day1–12 once, including idle days. Apply actual FIFO arrival admission and active WIP, then model build capacity with return-first order, then later-day quality checks by build day/ID. Preserve required B/E returns unless the supplied paired model prevents them; failed verification needs next-day return build and a later check. Newly completed work frees WIP the next day. outcomes:[{item,admittedAt,completedAt,returns,leadTime,cycleTime}] covers all six source items. Dates/times are null where unadmitted or incomplete; lead and cycle means have different start points. report:{completed,open,returns,buildUnits,verificationChecks,meanCompletedLeadTime,meanCohortLeadTime,meanCompletedCycleTime,leadReduction}. Reconcile actual trace and source demand. Completed means use only their observed population; cohort mean and benefit remain null when any outcome is open. Empty observed populations have null means. followup:{setupDay,inspectDay,owner,recommendation,nextAction,claim}: setup0, inspect13/14, owner developers, conditional-process-experiment. Respond to actual whole-cohort target with continue-pilot/continue-bounded-pilot or iterate/revise-flow-policy; any censoring requires extend/extend-observation. A short active cycle or selected completed subset cannot claim customer benefit. Trace/outcome row order may vary; within-day queue order is part of this declared model. Save partial progress and repair without Reset; real agreement, causation and approval remain pending.

What the checks verify

Your work is graded on 10 independent properties, not on matching one reference answer.

Policy, baseline, daily events and cohort results remain bounded, unique and finite.
The declared improvement hypothesis retains its observed root cause and target policy.
Historical full-cohort lead evidence and quality returns remain immutable.
Actual admission preserves arrival order and active WIP across unfinished quality and return work.
Actual build/check events preserve model capacity, queue order and required later-day quality work.
The complete observation window retains every active, idle and unsuccessful day.
Every source item's admission, completion, return count and censoring reflect actual replay consequences.
Observed populations, full-cohort benefit and work demand remain correctly distinguished.
Next learning responds to actual full-cohort target support or incomplete observation.
Prompt owned setup and post-observation inspection preserve conditional process learning.

Where this sits in the PSM-I blueprint

Domain: Developing People and Teams
Objective: Coaching and Mentoring
Skill: Coaching for Effectiveness

Part of PSM-I preparation

PSM-I

Professional Scrum Master I

All PSM-I practice All labs

Labs are written by ExamNova to teach the decisions the exam tests. They are not reproductions of vendor lab content.