<result> <column_name>AI_Usage_Hours_Per_Day</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.5</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 (a per-entity daily-rate level; summing across entities is not a natural rollup), product (not a compounding 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>The table is a DIMENSION with no sequence axis, so multiplicative is implausible. The "_Per_Day" token is the only ratio-like signal, and it is just a unit of hours per day, not a dimensionless factor around 1.0. The sample values (0.5-0.7) are hours with no negatives or zeros, so this reads as a per-entity daily usage level, where averaging is the natural rollup. The column is VARCHAR with dirty entries (an empty value, "0.6 hrs"), so numeric statistics are unavailable. The data needs cleaning and casting to numeric, and the subtype call is moderate-confidence.</reasoning> <table_type_conflict>false</table_type_conflict> </result>