After a digital change, a business may observe shorter time frames or fewer errors. Attributing all the change to the instrument, however, requires attention: in the same period volumes, people and complexity of the work can be changed. The comparison must be designed.
I propose to define first what will be measured and what activities will enter the test. An indicator such as average time per request can be affected by the presence of simpler cases. It is also useful to describe quantities, types and exceptions, without demanding greater precision than the available data.
A hypothetical example is a new function tested only by those who know the process very well. The result may not describe the difficulties of other users. A subsequent trial of relevant people and cases allows for necessary training and support to be recognized.
Verification should include work moved elsewhere: data preparation, corrections, and assistance. I propose to state what the comparison allows to conclude and what remains uncertain. Not every business test demonstrates a causal effect with experimental rigor; it can still provide useful information to decide whether to continue, modify or abandon the intervention before extending it.