Metrics
MMM (media-mix modeling)
Top-down statistical modeling of incremental contribution by channel.
MMM (media-mix modeling, sometimes 'marketing mix modeling') is the top-down statistical method of estimating each channel's incremental contribution to revenue using historical data. Pre-existed the cookie era, came back into fashion as bottom-up attribution decayed under signal-loss. Common tools: Robyn (Meta open-source), LightweightMMM (Google), Recast, Cassandra. Output: per-channel saturation curves and recommended budget allocations.
Example
DTC brand runs Robyn-based MMM on 18 months of weekly spend data — finds Meta is saturating at $40K/week with diminishing returns, while CTV and YouTube are still at the linear part of their curves — re-allocates $8K/week from Meta to CTV.