When extending small-sample results into a verification plan, define the experimental unit first. Repeating execution on one image, capturing the same workpiece again, repositioning it and using independent workpieces answer different questions. Record workpiece, image, capture event and configuration identifiers for each run.
Group records by factors that change
Organize verification by workpiece type, production batch, surface condition, pose, lighting or camera configuration. State which factors remain unchanged and which change. NIST data collection guidance records significant influencing factors. Use that recording approach. Agree on the sample groups and acceptance criteria for the target task.
Retain individual results and group summaries
Save expected results, actual outputs, timing and exceptions with their original samples. Report counts and decision differences by group, explaining repeated records and exclusions. Do not submit a summary percentage alone. Sample coverage and individual results show which conditions the statistical conclusion covers.
Check the affected scope after revisions
After changing parameters, reference images, models or image transformations, plan retests for the affected scope. Keep configurations and results from before and after the change. Run regression checks on related normal and abnormal samples. Include untested conditions and additional samples in the next plan.