Question

Statistical control is ...

1. Always effective

2. Good enough to prove causation

3. Only as good as your measurement

4. Sufficient enough to replace A/B testing and experiments

Answer 1

Statistical control is useful for understanding whether a process is operating within the desired limits. It helps to promptly identify in case there is anything wrong in the process.

However, even under statistical control, some errors of small magnitude may go undetected. Moreover, statistical control does not ensure a causal relationship. It can detect patterns in a process, such as in an experiment. However, it cannot replace hypothesis testing and experiments, as all of these have their own separate utilities- hypothesis testing helps to check whether the observed characteristics support the pre-conceived ideas and experiments help to establish or nullify hypothesized relationships.

If the measurements are appropriate, then statistical control also provides a reliable understanding of the accuracy of a process.

Hence, Statistical control is “3. Only as good as your measurement”.