Whichever type of data is being recorded, the frequency with which the samples are taken should be chosen in relation to the process itself.
For example, if the process is carrying out a task which takes all day, there is not much point in checking its outputs every hour. On the other hand, if a process is dependent on a series of prior events to keep it functioning properly, it would probably be sensible to take frequent samples, to spot a potential problem quickly.
In either case, the size of the sample and the frequency with which it is taken should be both straightforward and practical.
It is also recommended to monitor the process for a sufficiently long period of time before beginning to draw firm conclusions on how it is operating.
For example, when dealing with variable data, it is recommended that a minimum of 20 subgroups (comprising at least 100 individual items) be taken before any interpretation of the data is made.