Establish a baseline
Measure current task time, cost and quality. Gather data across several shifts, including busy periods.
Choose success metrics
For manipulation count correct picks and placements; for demonstrations count sessions matching the script. Always include failures, human interventions and downtime.
Test realistic conditions
Include people moving around, layout changes and real loads. A pilot in an empty hall answers fewer questions than a test during normal operation.
Record the decision and next step
Compare the result with the threshold agreed before the trial. Record what works, what needs changing and the cost of expansion. Do not scale up before resolving the main limitations.