Zero-Rating, One Big Mess: Analyzing Differential Pricing Practices of European MNOs

TL;DR

This study tackles the opacity of zero-rating practices by EU MNOs, auditing how traffic from four popular apps is classified across seven operators using a controlled roaming-focused testbed. The authors develop a methodology combining market analysis, a three-experiment measurement framework, and the MobileAtlas platform to reveal whether operators apply IP-based or hostname-based classification and to detect misclassifications and roaming-coverage gaps. Key findings include pervasive misclassification, uneven handling of IPv6 and HTTP/3, and partial roaming-zero-rating adherence, highlighting transparency gaps and regulatory concerns. By open-sourcing their experiments and infrastructure, the work enables replication and supports policymakers, regulators, and researchers in auditing differential pricing practices.

Abstract

Zero-rating, the practice of not billing data traffic that belongs to certain applications, has become popular within the mobile ecosystem around the globe. There is an ongoing debate whether mobile operators should be allowed to differentiate traffic or whether net neutrality regulations should prevent this. Despite the importance of this issue, we know little about the technical aspects of zero-rating offers since the implementation is kept secret by mobile operators and therefore is opaque to end-users and regulatory agencies. This work aims to independently audit classification practices used for zero-rating of four popular applications at seven different mobile operators in the EU. We execute and evaluate more than 300 controlled experiments within domestic and internationally roamed environments and identify potentially problematic behavior at almost all investigated operators. With this study, we hope to increase transparency around the current practices and inform future decisions and policies.

Figures (4)

• Figure 1: Architecture and components of the MobileAtlas measurement platform
• Figure 2: Involved actors and traffic flow when verifying zero-rated data traffic
• Figure 3: Actors and traffic flow when checking for IP-based classification
• Figure 4: Involved actors and traffic flow when checking for hostname-based classification