Yesterday I swapped one word in a subhead and Parse.ly showed engaged time inching from 2:41 to 2:49 across 1,842 visits, with read-through up 0.6 percentage points — real lift or just noise? For folks A/B testing articles, what’s your benchmark to call it a win (engaged time, read-through, CTR), and how big a delta at about 2k sessions feels worth shipping?