What Is Shapley Attribution?
Shapley attribution applies the Shapley value from cooperative game theory to assign fractional conversion credit across the touchpoints in a user journey. Each touchpoint is treated as a player in a cooperative game where the payoff is the conversion. The Shapley value is the unique credit assignment that satisfies four fairness axioms: efficiency (credits sum to one), symmetry (interchangeable touchpoints get equal credit), null player (touchpoints that add no value get zero), and additivity (multi-game results compose).
In marketing, Shapley produces a per-touchpoint credit allocation by averaging that touchpoint's marginal contribution across all possible permutations of the journey.
Why Shapley Matters
Shapley is the theoretically rigorous answer to multi-touch attribution. The fairness axioms it satisfies are the desirable properties any reasonable credit allocation should have. Rules-based MTA (linear, time-decay, U-shaped) violates at least some of these axioms; Shapley by construction does not.
Practically, Shapley is the default underneath Google Ads data-driven attribution and several enterprise attribution platforms. When a vendor says "algorithmic attribution" they usually mean Shapley or a close variant (Markov chains is the other common choice).
How Shapley Attribution Works
For each user journey, compute the touchpoint's marginal contribution: the conversion probability with the touchpoint minus the conversion probability without it, averaged across every possible ordering of the journey's other touchpoints. This is the Shapley value for that touchpoint in that journey. Aggregate across users to get channel-level credit shares.
The computation is expensive (factorial in the number of touchpoints) and is approximated in practice via sampling or by exploiting structure in the journey graph. Google Ads, Adobe, and most algorithmic attribution vendors use sampling-based approximations.
Limitations for AI Search
Shapley is rigorous only over the touchpoints in the journey. Touchpoints not in the data, including all AI assistant interactions, get zero credit by construction. Shapley does not solve the AI search blind spot; it solves the credit-distribution problem within touchpoints the system can already see. For AI search, the answer remains MMM with an AI visibility proxy.
The risk is that "we use Shapley" can be presented as if it addresses the AI attribution gap. It does not. The vendor's Shapley model is more rigorous than rules-based MTA for tracked channels but no more capable of seeing AI.
How Presenc AI Helps
Presenc AI provides the AI visibility data that complements any Shapley-based attribution: feed the visibility series into MMM as a media-equivalent variable, run Shapley for within-channel tactical optimization in paid digital, and triangulate. The combination produces both within-channel rigor and cross-channel completeness.