A metric moved four percent this quarter and the team reads it two ways. The report needs one headline — which reading do we lead with?

The single headline should lead with composition: most of the four percent move is driven by account mix, specifically new accounts that score lower in every period on record. That pattern is stable and structural/mechanical, so it is the dominant explanation of the overall change and the safest basis for a report headline. Leading with behaviour would elevate a smaller within-cohort residual above the main driver. The within-cohort movement is visible and worth monitoring or investigating as a follow-up caveat, but it is not strong enough to headline as a behaviour shift. Making behaviour the headline risks the classic Simpson's-paradox misread: treating a mix artifact as a genuine performance change, over-alarming readers, and misdirecting attention or resources toward the wrong intervention. The prudent conclusion is therefore to lead with composition, wait a quarter before declaring a behaviour shift, and keep the smaller cohort-level signal under active review rather than discarding it. This preserves report integrity by ensuring the one headline explains what actually moved while still acknowledging the secondary behavioural evidence.