Multi-touch attribution (MTA)
Splitting one conversion's credit across the ads, emails, searches, and links that appeared along the path, not just the last one.
See it
What it is
Multi-touch attribution splits one conversion's credit across the recorded touchpoints that came before it instead of awarding everything to one. A linear model shares credit evenly, time decay gives more to recent touches, and position-based models favor the first and last while leaving some credit for the middle. Note that Google retired those rule-based models in Google Ads and GA4 during 2023, leaving data-driven and last click, so you may have to build the split yourself in a warehouse.
Reach for MTA when journeys regularly span several ads, emails, searches, or referrals and last click keeps making the closing channel look like it worked alone. It gives a consistent accounting view for comparing assists, provided every report uses the same identity rules, attribution window, and model.
Gotcha: the model can only split credit across touchpoints it observed. Cookie loss, cross-device use, offline contact, and closed ad platforms leave holes that polished decimals cannot repair. MTA also describes association, not incremental lift. A channel can receive credit on every path while causing none of the conversions.
Ask AI for it
Build a BigQuery multi-touch attribution query from touchpoints(user_id, touch_id, channel, touched_at) and conversions(user_id, conversion_id, value, converted_at). Limit eligible touches to the 30 days before each conversion. Calculate linear, 7-day half-life time-decay, and 40/20/40 position-based credit with SQL window functions. For position-based only, use 50/50 when a path has exactly two touches; for every model, give a single-touch path 100%. Return channel credit and revenue by model, and add assertions that each conversion's weights sum to 1.0 within a tolerance of 1e-9.