When a central bank says it is data-dependent, how do you test whether the data is actually driving decisions?

Central banks often describe their policy as data-dependent, but that phrase is hard to falsify. Almost any decision can be described afterwards as a response to the latest print, so the label alone tells us little. One way to make it testable is to compare what the bank said it was watching against what it did. If the stated trigger was core services inflation, did the decision change when that series moved, and did it stay put when headline inflation moved for reasons the bank had called transitory? Looking at the minutes and the forecast revisions over several meetings can show whether the reaction function is consistent or whether the goalposts shift. I'm not certain how much of this can be done cleanly from public material, since minutes are edited and projections are published with their own caveats. I'd be interested in how others here separate a genuine change in the reaction function from a rationalisation after the fact. What evidence would convince you that a bank's stated data focus was a poor guide to its actual decisions?