Towards a Fair Allocation of Rewards in Multi-Level Marketing

Talal Rahwan, Victor Naroditskiy, Tomasz P. Michalak, Michael J. Wooldridge, Nicholas R. JenningsarXiv 2014 · arXiv

Introduction

Multi-level marketing (MLM) and referral programs are foundational to modern growth strategies, yet designing a reward system that is both fair and effective remains a challenge. Traditional models often fail to balance the incentives between the direct referrer and the chain of "ancestors" who made the referral possible. This paper addresses this gap by applying the Shapley value—a foundational fairness concept from cooperative game theory—to a new class of "tree games." The authors demonstrate that for basic referral structures, this theoretically optimal reward distribution can be computed efficiently in linear time.

The Fairness Gap in Current Models

Existing industry standards typically fall into two categories, each with distinct drawbacks:

  1. The "Refer-a-friend" Model (e.g., Dropbox): This model rewards both the referrer and the invitee equally. While simple, it ignores the contributions of ancestors further up the chain, failing to incentivize the creation of deep referral networks.
  2. Geometric Reward Schemes: Common in challenges like the DARPA Red Balloon Challenge, these split rewards among ancestors in decreasing shares (e.g., 1/2 to the parent, 1/4 to the grandparent). This often leaves the newest participant with little to no immediate reward, potentially discouraging entry.

The goal of this research is to find a mathematically grounded "middle ground" that rewards both direct and indirect contributions while maintaining computational feasibility.

Methodology: Tree Games and the Shapley Value

The researchers model referral marketing as a Cooperative Game, specifically defining a subclass called Tree Games. In this framework, agents are nodes in a rooted tree. A coalition of agents only generates value if they form a connected path back to the "root" (the company).

Illustration of a tree $T$, an induced sub-graph $T^C$ based on a coalition $C$, and the resulting trimmed tree $T^C_{trim}$. This process is central to defining the value of coalitions in tree games.

To determine fair payoffs, the authors utilize the Shapley value, which calculates a player's reward by averaging their marginal contribution across all possible joining sequences. While calculating the Shapley value is generally NP-complete, the authors prove that for "Basic Tree Games"—where each referral has a fixed value—the distribution follows a simple formula and can be calculated in $O(n)$ linear time.

Key Insights: The "Equal Share" Rule

A significant finding of the paper is the emergence of the "Equal Share" rule for basic referral trees. Under the Shapley mechanism, the value generated by a single referral is distributed in equal shares among the invitee and all their ancestors in the chain.

This approach acknowledges the indirect value of ancestors; without the "great-grandfather" of a chain, the newest member would never have been reached. Furthermore, the authors prove these games are convex, meaning the Shapley value lies within the "core"—a state where no group of players has an incentive to defect, ensuring the stability of the marketing network.

Comparative Results

The paper evaluates the Shapley mechanism against existing models using a sample referral tree. The results highlight how the Shapley value balances the extremes of other methods.

Comparison of reward allocations across different models for a sample referral tree. The Shapley value provides a mathematically justified middle ground.

As shown in the table, while the Geometric model may leave leaf nodes (e.g., nodes 6 and 7) with zero reward, and the Dropbox model provides minimal incentive to the root (node 1), the Shapley value ensures every participant in the chain receives a non-zero, fair portion of the total value created.

Conclusion and Future Directions

While this work provides a robust theoretical foundation for fair MLM rewards, several avenues for future research remain. The current model assumes a strict tree structure, whereas real-world social networks are complex graphs where individuals may receive multiple invites. Additionally, the model does not yet account for the varying "costs" or effort levels associated with making a referral. Empirical validation is also necessary to determine if users perceive the Shapley distribution as fairer in practice, potentially leading to higher engagement and retention in referral programs.

Summary updated

Abstract

An increasing number of businesses and organisations rely on existing users for finding new users or spreading a message. One of the widely used "refer-a-friend" mechanisms offers an equal reward to both the referrer and the invitee. This mechanism provides incentives for direct referrals and is fair to the invitee. On the other hand, multi-level marketing and recent social mobilisation experiments focus on mechanisms that incentivise both direct and indirect referrals. Such mechanisms share the reward for inviting a new member among the ancestors, usually in geometrically decreasing shares. A new member receives nothing at the time of joining. We study fairness in multi-level marketing mechanisms. We show how characteristic function games can be used to model referral marketing, show how the canonical fairness concept of the Shapley value can be applied to this setting, and establish the complexity of finding the Shapley value in each class, and provide a comparison of the Shapley value-based mechanism to existing referral mechanisms.
BibTeX
@misc{rahwan2014towards,
  author = {Talal Rahwan and Victor Naroditskiy and Tomasz P. Michalak and Michael J. Wooldridge and Nicholas R. Jennings},
  title = {{Towards a Fair Allocation of Rewards in Multi-Level Marketing}},
  year = {2014},
  eprint = {1404.0542},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1404.0542},
}
Figure