C.16.RM:8 - Common Anti-Patterns and How to Avoid Them

Failure in these situations	Consequence	Repair
Repeat a mixed signal and average it	The stable background remains in the result.	Cancel, estimate or control the consequential contribution through a justified relation.
Add an unknown correction without a way to bound or determine it	The formula changes while the target stays unresolved.	Obtain the missing relation or choose an arrangement that limits the influence.
Improve agreement by changing the wanted subject condition	The new reading answers a different question.	Preserve the target or derive the relation connecting the changed condition to it.
Treat an implementation test as confirmation of the physical model	Correct arithmetic is used to support an untested subject premise.	Match the comparison to the contribution whose adequacy is in question.
Continue fault diagnosis after a sufficient repair	The present result is delayed for an unrelated explanation.	Return the adequate result and pursue the remaining cause only for a question that needs it.
Apply a new correction to an old result without checking its conditions	The revised value can misrepresent the earlier measurement.	Reuse the retained observations through a relation supported for that situation.