Attribution
Multi-touch attribution
Also called MTA.
Multi-touch attribution is any model that splits credit across more than one touch on the path to conversion. It includes linear, time-decay, position-based, and data-driven models.
How it is measured
The system builds a path of touches for each conversion within the lookback, applies the model's weights, and sums the credit per channel. The totals add back up to the conversion value.
Check the path coverage. A model can only share credit among touches it can see. Offline touches and untagged links are missing, which pushes credit toward visible channels.
Worked example
A boutique tea company reviews 135 orders from 24 days. Paths include a podcast mention, a search click, a newsletter, and a marketplace link. 68 paths have two or more tracked touches.
Under last-click, the marketplace link gets 31 percent of credit. Under a multi-touch split it gets 19 percent while the newsletter rises from 12 to 22. The company moves effort to a welcome email flow.
How it differs
Multi-touch attribution is the family. Linear attribution is one fixed rule inside it that splits evenly. Choose the family when you care about the path; choose a member when you want a rule.
Common errors
Assuming more touches means more truth. Ignoring untracked touches. Using the model to micromanage channels. Mixing it with last-click totals in one table. Skipping a sanity check against a test.
In practice
Report one multi-touch model next to last-click for a quarter. List which channels change rank. Use differences as questions to test, not as orders.