[Paper Review] Revenue-based Attribution Modeling for Online Advertising

Kaifeng Zhao, Seyed Hanif Mahboobi+1 more|arXiv (Cornell University)|Oct 18, 2017

Consumer Market Behavior and Pricing13 references3 citations

TL;DR

This paper proposes revenue-based attribution modeling using relative importance methods—dominance analysis and relative weight analysis—applied to regression models to fairly allocate revenue contributions to online advertising channels. It extends these methods to additive models, demonstrating superior accuracy and flexibility over traditional approaches in simulations and real-world data.

ABSTRACT

This paper examines and proposes several attribution modeling methods that quantify how revenue should be attributed to online advertising inputs. We adopt and further develop relative importance method, which is based on regression models that have been extensively studied and utilized to investigate the relationship between advertising efforts and market reaction (revenue). Relative importance method aims at decomposing and allocating marginal contributions to the coefficient of determination (R^2) of regression models as attribution values. In particular, we adopt two alternative submethods to perform this decomposition: dominance analysis and relative weight analysis. Moreover, we demonstrate an extension of the decomposition methods from standard linear model to additive model. We claim that our new approaches are more flexible and accurate in modeling the underlying relationship and calculating the attribution values. We use simulation examples to demonstrate the superior performance of our new approaches over traditional methods. We further illustrate the value of our proposed approaches using a real advertising campaign dataset.

Motivation & Objective

• To develop a more accurate and flexible method for attributing revenue to online advertising channels.
• To address limitations of traditional attribution models that rely on simplistic or arbitrary allocation rules.
• To extend relative importance techniques from linear to additive regression models for better modeling of complex advertising effects.
• To provide a data-driven, statistically sound framework for measuring the true contribution of each advertising input to overall revenue.

Proposed method

• Adopt and extend the relative importance method to decompose the R² of regression models into contributions from individual predictors (advertising channels).
• Implement two submethods: dominance analysis, which evaluates average marginal contributions across all subsets of predictors, and relative weight analysis, which uses orthogonal components to assess relative importance.
• Apply the decomposition techniques to both standard linear models and additive models to capture non-linear relationships between advertising inputs and revenue.
• Use simulation studies to validate the performance of the proposed methods against traditional attribution models.
• Apply the methods to a real advertising campaign dataset to demonstrate practical utility and robustness.
• Ensure interpretability and statistical rigor by grounding attribution values in regression theory and variance decomposition.

Experimental results

Research questions

• RQ1 How can we fairly attribute revenue to individual online advertising channels in a statistically principled way?
• RQ2 What are the limitations of traditional attribution models in capturing complex, non-linear relationships between advertising efforts and revenue?
• RQ3 Can relative importance methods like dominance analysis and relative weight analysis provide more accurate and flexible attribution than standard approaches?
• RQ4 How do these methods perform when extended from linear to additive models in capturing non-linear effects?
• RQ5 To what extent can these methods improve revenue attribution accuracy in real-world advertising datasets?

Key findings

• The proposed methods—dominance analysis and relative weight analysis—demonstrate superior performance in accurately attributing revenue compared to traditional attribution models in simulation studies.
• Extending the decomposition methods to additive models enhances flexibility and improves modeling of non-linear relationships between advertising inputs and revenue.
• The relative importance framework provides a statistically sound basis for allocating revenue contributions, reducing bias from arbitrary or heuristic allocation rules.
• Empirical results on a real advertising campaign dataset confirm the practical value and robustness of the proposed approach in real-world settings.
• The methods are shown to be more reliable and interpretable than conventional approaches, especially in scenarios with correlated advertising channels.
• The study confirms that revenue-based attribution using relative importance yields more accurate and actionable insights for digital advertising optimization.