<result> <column_name>build_ms</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>mean</primary_reducer> <valid_reducers>mean, median, min, max, p50, p95, p99, distribution, last</valid_reducers> <excluded_reducers>sum (summing durations across time steps 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. The column name "build_ms" indicates a duration measurement in milliseconds - an instantaneous physical quantity at each sequence step. The large magnitude (max ~2.5M ms), wide spread (std ~335K), and center far from both 0 and 1 (mean ~83K, median ~1K) rule out multiplicative. This is clearly additive. The critical distinction: on a PATH table, instantaneous durations are ordinal_additive - summing build times across timestamps is meaningless (you don't add "10ms at t1" + "20ms at t2" = "30ms total"). The natural rollup is mean/median/percentile across entities at each step, or last/distribution along the sequence. Classified as ordinal_additive with mean as primary reducer. </reasoning> <table_type_conflict>false</table_type_conflict> </result>