When you only check 1 thing at a time and you see a transparent distinction in the metric you’re analyzing, you presumably can draw an correct conclusion. However, if you had also modified the sender’s name, it would be impossible to conclude that the topic line made all
