When a CPI print and its prior-month revision point different ways, what do you check first?

A recurring puzzle in macro reading is a consumer price release where the headline monthly change looks hot but the revised prior month came in softer, or the reverse. The headline is usually the one that moves markets on the day, yet the revision changes the trend picture, and the two can tell different stories about whether inflation is accelerating or cooling. One way to separate them is to look at the core component against the headline, then at the three- and six-month annualised rates, which smooth out single-month noise. Energy and food swings can dominate the headline in a given month, so the detail often matters more than the top-line number. Comparing against the release calendar and the period being measured also helps, since a print can be dated to a different reference month than the one a casual reader assumes. I am not certain how consistently members weigh revisions against the first estimate, and I would not assume there is a single correct convention across institutions. What do you look at first when the headline and the revision disagree, and why?