When a CPI print surprises, how do you separate the headline from the core read?

A headline CPI number can move rates and FX on its own, but the component detail often tells a different story. Energy and food swings can push the headline well away from core, and the first-release figure may later be revised when seasonal factors or late price collections come through. Worth checking which month the print actually refers to before reacting to the month-on-month and year-on-year numbers, since they can point in different directions. Some people I follow on this topic lean on the three-month annualised core rate or services ex-shelter as a cleaner gauge of persistence. Others argue the market reacts mainly to the surprise versus consensus, so the level matters less than the gap. Both views seem reasonable depending on the regime, and I am not certain which one has held up better across recent cycles. What is your process for deciding whether a surprise changes the underlying trend, and which component or revision do you check first?