Reference library · Digital attribution

Multi-Touch Attribution (MTA)

Multi-Touch Attribution (MTA) is a measurement method that splits conversion credit fractionally across every touchpoint in a user's path to conversion, instead of crediting one channel entirely. Credit is distributed by a weighting model, such as linear, time-decay, position-based, or algorithmic, applied to touchpoint-level path data stitched together at the user or device level. The output is a set of channel-level credit shares used to compare channels on a like-for-like basis rather than by last click alone.

Formula

Touchpoint credit = decay weight_i / sum of all decay weights x conversion value, where decay weight_i = 0.5^(days before conversion / half-life)

Worked example

A €120 subscription purchase follows four touchpoints: a display ad seen 10 days before conversion, a Facebook ad 5 days before, a Google Search ad 2 days before, and a direct site visit on the conversion day itself. Using a time-decay model with a 7-day half-life, decay weight = 0.5^(days before conversion / 7): display = 0.372, Facebook = 0.610, Google Search = 0.820, direct = 1.000, summing to 2.801. Dividing each weight by 2.801 and multiplying by €120 gives credit of €15.91 to display, €26.11 to Facebook, €35.14 to Google Search, and €42.84 to direct, which sum back to €120.00. Search and the direct visit near conversion take most of the credit, but display's €15.91 confirms it earns a place in the plan even though last-click reporting would give it nothing.

How it is used

MTA output feeds channel-level budget reallocation, campaign bidding, and path-length analysis for creative sequencing across the funnel. Model choice changes which channels look strong, so planners should run at least two models (e.g. linear and time-decay) and check whether the ranking of channels holds before committing budget shifts. The practical failure mode is running MTA on incomplete paths: Safari ITP, Chrome's cookie phase-out, and iOS ATT block cross-site tracking for a growing share of users, so paths get truncated and credit inflates toward the last visible touch (branded search, direct, retargeting) while upper-funnel display and video are starved of credit.

The common mistake

Confirm any MTA-driven budget shift with a holdout or geo lift test, since the credit split reflects the model's correlation, not measured incrementality.

← All terms Calculated in ValueMetrics →