How long should an observation sitting last?
Teams sometimes ask us to “just watch for a full day.” A full day produces notebooks full of noise and tired operators who perform for the stranger in the chair. We prefer deliberate lengths.
Forty to sixty minutes suits a single workflow review: enough to see setup, a few repetitions, and one recovery from error. Ninety minutes fits consoles with long-running cases that span several screens. Multi-hour observation is reserved for fatigue and interruption studies — and only with clear consent and breaks.
What matters more than clock time is task coverage. A short sitting that includes the dreaded edge case beats a long sitting that only repeats the happy path. Our observation plans list the tasks we must see at least once; if a sitting ends before that list is met, we schedule another rather than invent findings.
If you run internal observation without us, decide the length before you sit down, tell the operator, and stop when you said you would. Respect is part of the data quality.