<result> <column_name>confidence</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.92</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, min, max, median, distribution</valid_reducers> <excluded_reducers>sum (summing confidence scores is meaningless), product (not a ratio)</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> Gate prior: DIMENSION table with no sequence axis makes multiplicative highly implausible. The column name "confidence" indicates a probability or certainty score - an intensive level quantity bounded in [0,1]. While the median (0.99) and mean (0.89) are near 1.0, this is a bounded probability scale, not a ratio centered on 1.0 as a "no change" value. The presence of zeros (0.17%) and the bounded [0,1] range confirm this is an absolute level, not a multiplicative factor. Summing confidence scores across entities is meaningless; averaging them is the natural aggregation. Classified as intensive_additive with high confidence. </reasoning> <table_type_conflict>false</table_type_conflict> </result>