<result> <column_name>total_rows</column_name> <family>additive</family> <subtype>ordinal_additive</subtype> <confidence>0.85</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>last</primary_reducer> <valid_reducers>last, max, mean, median, min</valid_reducers> <excluded_reducers>sum (summing snapshot counts across time 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: PATH table with sequence axis allows full analysis. Column name "total_rows" indicates a count/snapshot level. The values (782 to 307957) are large absolute counts, not ratios near 1.0, ruling out multiplicative. The sequence column "deleted_at" suggests this tracks row counts at deletion events - a snapshot level at each sequence step. Summing snapshot counts across time is meaningless (you don't add "782 rows at time 1" to "3614 rows at time 2"), but the last value or max/mean are sensible aggregations. This is ordinal_additive: additive across entities at a fixed sequence step, but only ordinally meaningful along the sequence axis. </reasoning> <table_type_conflict>false</table_type_conflict> </result>