Work sampling: activity classes and a random observation plan

Why it matters

If the question is how common waiting is in a work area, it may not be necessary to follow the start and end of every job. Work sampling records the state of the area at moments chosen in advance. Each record is a short cross-section; it does not show how all the time in between was spent. When many properly chosen cross-sections are evaluated together, the share of time spent in each activity is estimated. The unit of measurement here is not the second but one observation moment for a defined person or resource. Seeing two different resources at the same time does not mean two independent moments were drawn. Increasing only the number of records without stating the target quantity can mean collecting more data for the wrong question.

Seeing an activity in 30% of the observation moments does not prove that 30% of every work cycle consists of that activity. Among long and scattered jobs, even the cycle boundary may not be shared. A time share is interpreted for the time universe that was selected. Whether a worker who is waiting is doing another necessary job depends on the class definition. Pooling a human activity and a machine state in the same column hides what the person does while the machine runs. First you must choose whom or what you are observing. That way the person waiting, the machine running and the part sitting in a queue remain separate questions; the proportion of one is not turned into the duration of another.

Random selection does not mean the observer goes to the area at moments that suit the observer. The selection must rest on a rule that covers all the eligible times being targeted and is applied without looking at the result. Moving records to calm hours because observing at busy moments is hard ties the chance of being selected to the very event under study. A bigger record table does not remove this link. For the same reason, choosing only the resources close to the walkway may not represent the whole area. Accessibility is a condition that has to be resolved while the observation plan is prepared. Reporting afterwards only on the part that is easy to observe is possible, but that result is not the answer to the broad target set at the start.

When activity classes are clear, different observers can record the same moment in a similar way. But writing the names of the classes is not enough; boundary cases need decision rules. If K01 cleans a tool while waiting for a part, the moment falls into the cleaning class when human activity is under study. If the view is blocked, it is not assigned to the waiting class. A moment that cannot be seen is a missing observation and is kept apart from the activity totals. The purpose of the measurement is not to make the person faster or to demand uninterrupted motion. It is to make the conditions of waiting visible and to choose the right improvement question. Used together with the method study in M3, this approach lets you question not only the duration of an unnecessary activity but also why it exists.

Formula