Algorithmic Multi-Touch Attribution: Implementing Shapley Value Models in Python

Last-touch attribution is the primary driver of wasted commercial capital in enterprise tech. By crediting 100% of a multi-million-dollar deal to the final demo request form or branded search click, marketing teams defund the high-value technical architecture content that actually built consensus across the buying committee. Cooperative game theory offers a mathematically sound alternative: Shapley Value Attribution.

1. The Cooperative Game Theory Formulation

In cooperative game theory, the Shapley Value assigns a unique distribution of total surplus generated by a coalition of players. In revenue engineering, the coalition consists of the marketing channels that touched an enterprise account, and the surplus is the closed-won ARR.

2. Production Python Implementation

import itertools
from collections import defaultdict

def compute_shapley_attribution(journeys, revenue_dict):
    channels = sorted(list(set([ch for j in journeys for ch in j])))
    n = len(channels)
    shapley_values = defaultdict(float)
    
    for channel in channels:
        other_channels = [c for c in channels if c != channel]
        for r in range(len(other_channels) + 1):
            for subset in itertools.combinations(other_channels, r):
                subset = list(subset)
                subset_with = tuple(sorted(subset + [channel]))
                subset_without = tuple(sorted(subset))
                
                v_with = revenue_dict.get(subset_with, 0.0)
                v_without = revenue_dict.get(subset_without, 0.0)
                marginal_contribution = v_with - v_without
                
                weight = (math.factorial(len(subset)) * math.factorial(n - len(subset) - 1)) / math.factorial(n)
                shapley_values[channel] += weight * marginal_contribution
                
    return dict(shapley_values)

3. Transforming Attribution Insights into Budget Allocation

When marketing architects analyze pipeline with game-theoretic models, high-information SEO moats and generative AEO citations regularly claim 40%+ of attribution credit, proving that technical content drives the lion's share of commercial enterprise valuation.