<result> <column_name>BOUNCE RATE (%)</column_name> <family>uncertain</family> <subtype>none</subtype> <confidence>0.4</confidence> <requires_user_confirmation>true</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, median, min, max, distribution</valid_reducers> <excluded_reducers>sum (a percentage level does not accumulate across entities), product (values are on a 0-100 scale, not compounding factors near 1.0)</excluded_reducers> <average_semantics>arithmetic_mean</average_semantics> </reducers> <inner_outer_measure> <inner_measure_reducer>min</inner_measure_reducer> <outer_measure_reducer>max</outer_measure_reducer> </inner_outer_measure> <multiplicative_partner_derivable>false</multiplicative_partner_derivable> <reasoning>The table is STATE with no sequence axis, so the prior strongly disfavors multiplicative. The name ("rate", "%") is the only multiplicative evidence. The data contradict it: the median is 0, 57.7% of values are zero, the range is 0-100, and the mean is about 16, far from 1.0. There is no sequence for cumprod or cumsum reconstruction, so nothing verifies a compounding ratio. The column behaves like a bounded percentage level, which would be summed never and averaged (ideally weighted by sessions), but naming and numeric signals conflict, so I abstain with the family as uncertain and flag it for human review.</reasoning> <table_type_conflict>false</table_type_conflict> </result>