Multi-Touch Attribution (MTA) distributes credit for a conversion across the sequence of marketing touches that preceded it, instead of crediting one touch (first-click, last-click). MTA models range from rule-based (linear, time-decay, position-based) to data-driven (Markov chains, Shapley values).
For enterprises, the warehouse-MTA pattern (deterministic per-user touch sequences in Snowflake / BigQuery) is the closest thing to ground truth for cross-channel attribution. It almost always disagrees with platform-reported attribution -- Google Ads, Meta, Amazon all over-claim their own contribution.
The discipline: use platform attribution for in-platform bid optimisation, use GA4 for cross-channel diagnostic, use warehouse-MTA / MMM for cross-channel budget allocation. Every dashboard names which model it's reporting, so "5x ROAS in Google Ads" and "3.2x marginal ROAS in MMM" are both true and comparable.