Optimal Pricing for Peer-to-Peer Sharing With Network Externalities

Yunpeng Li, Costas Courcoubetis, Lingjie Duan, Richard R. WeberIEEE/ACM Trans. Networking 2021 · DOI

TL;DR

This paper tackles the challenge of setting optimal prices for peer-to-peer sharing platforms, where each new user adds value to existing users (positive network externalities). They introduce a novel "Price of Information" metric to quantify profit loss due to incomplete user knowledge and propose a simple, asymptotically optimal uniform pricing strategy that can replace complex differentiated pricing.

What Problem Does This Paper Address?

Peer-to-peer (P2P) sharing platforms, from WiFi sharing to social media, are booming. A key characteristic of these platforms is the presence of positive network externalities: the more users join, the more valuable the service becomes for everyone. This creates a complex pricing problem for platform operators who want to maximize profit. They need to decide who to let join and how much to charge, all while accounting for the fact that users' willingness to pay (their "valuation") is often private information. Existing approaches often assume complete information or rely on complex pricing schemes. This paper aims to bridge this gap by analyzing optimal pricing under incomplete information and proposing a practical, simplified pricing strategy.

Methods

The authors model two types of P2P sharing: bounded utility (like WiFi coverage, where value plateaus) and unbounded utility (like social media, where value grows with users, following Metcalfe's or Zipf's laws). They analyze pricing strategies under both complete and incomplete information about user valuations.

A core contribution is the introduction of the Price of Information (PoI), a ratio comparing profits under complete versus incomplete information. This metric quantifies the profit loss due to not knowing users' exact valuations.

For incomplete information, they first derive an "optimal/differentiated pricing scheme" which, while theoretically optimal, is complex. They then propose a uniform pricing scheme as a practical approximation. A key finding is that this uniform pricing scheme is asymptotically optimal, meaning it performs nearly as well as the complex differentiated scheme as the number of users grows large.

The paper also extends the analysis to a two-sided market, where some users contribute to the network (like WiFi sharers) and others only consume.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Price of Information (PoI): This new metric provides a clear way to measure the economic impact of imperfect information on platform profits. It's analogous to the "Price of Anarchy" but focuses on profit loss rather than social welfare.
  • Uniform Pricing is Surprisingly Effective: Even with incomplete information, a single, uniform price can achieve near-optimal profits as the user base grows. This significantly simplifies platform management compared to complex, user-specific pricing.
  • Less Concave Utility = Higher PoI: The profit loss due to lack of information (PoI) is greater when the network externality function is less concave. This means platforms with rapidly growing value (like social media) are more sensitive to information asymmetry than those with diminishing returns (like WiFi coverage).
  • Zero Pricing for Contributors in Two-Sided Markets: To attract "pure" consumers in a two-sided market, platforms might strategically charge zero to the users who contribute to the network's value.

Key Results

  • For bounded utility models, the PoI is exactly 2, meaning the platform loses half its potential profit due to incomplete information.
  • For unbounded utility models, the PoI is in the range [2, 27/8], with higher values for less concave utility functions (e.g., Metcalfe's law yields a higher PoI than Zipf's law).
  • The uniform pricing scheme is shown to be asymptotically optimal for both bounded and unbounded utility models, meaning its profit converges to that of the optimal differentiated scheme as the number of users approaches infinity.
  • In a two-sided market, the optimal pricing strategy can involve charging zero to contributors to attract consumers, and the PoI generally increases as the proportion of contributors decreases.

Limitations and Future Work

  • Assumptions: The analysis relies on specific assumptions about user valuation distributions (e.g., uniform distribution) and the functional forms of network externalities (Metcalfe's, Zipf's, or linear). Future work could explore more general distributions and externality functions.
  • Dynamic Pricing: The current model focuses on static pricing. Real-world platforms often adjust prices dynamically. Extending the model to dynamic pricing would be a valuable next step.
  • Strategic User Behavior: While incentive compatibility is considered, more complex strategic interactions among users (beyond simply reporting their valuation) could be explored.
  • Cost Structure: The paper assumes a simple per-user cost. More nuanced cost structures could be investigated.

Summary updated

BibTeX
@article{li2021optimal,
  author = {Yunpeng Li and Costas Courcoubetis and Lingjie Duan and Richard R. Weber},
  title = {{Optimal Pricing for Peer-to-Peer Sharing With Network Externalities}},
  journal = {IEEE/ACM Trans. Networking},
  year = {2021},
  doi = {10.1109/tnet.2020.3029398},
}
Figure