> Completed synthetic counterpart: `virtual-validation.html`, regenerated by `python run.py`. `completion.json` records the executed cases. The physical procedure below is retained for a real installation; it is not presented as unfinished computational work or as an executed physical test.

# Inspection time study

The supplied workloads are synthetic. Collect a representative sample before using the result for staffing decisions.

1. Define the boundary: when a job is offered, admitted, starts inspection, leaves inspection, returns for rework and finally leaves the cell. Decide whether jobs refused by the cell wait upstream, are diverted or are canceled. Record their fate.
2. Capture at least several complete shifts across the known product mix and demand conditions. Include beginning and end WIP, scheduled and actual breaks, staff availability, equipment downtime and incomplete jobs. Continue across shifts if carryover is normal.
3. Use one anonymous job ID across every visit. Record arrival, admission, each inspection start/end, reason for return and final disposition. Timestamp resolution should be much smaller than the shortest observed service duration. Keep the same time zone and synchronized clock.
4. Estimate interarrival behavior by time of shift and product class. Check batch arrivals, correlation and bursts before assuming a Poisson process. Separate active inspection duration from waiting, travel and rework delay.
5. Fit or resample service time within meaningful classes. Check sample size, outliers and missing records. Never infer rejected or unobserved jobs from completed work alone.
6. Reserve complete shifts for validation. Compare distributions of arrivals, throughput, waiting and unfinished work. Record where the model disagrees before choosing an operating policy.
7. Test an approved staffing/buffer change on a limited trial. Record labor time, floor-space constraints, upstream queueing, quality escapes and on-time completion. Costs and quality outcomes are not included in this numerical package.

`time-study-template.csv` contains a header only, not fabricated observations. Use a row for each inspection visit. For a job that never enters, leave service fields blank and state the upstream disposition. Do not place names or confidential production details in a shared copy.