It does not store real consumer data nor imposes any uniform metric; rather, it provides industry-proposed common code components that local implementers can choose to leverage as they seek to align with the advertisers’ ‘North Star’ approach to cross-media measurement.
Framework
Framework & Code Components
The Halo Framework is at the core of this voluntary initiative: a set of free-to-use methodological principles and privacy-preserving technologies developed under the Apache 2.0 open-source license.
Key Technological Pillars
Virtual People Framework (VPF)
- A distributed identification model combining panel data and digital logs into a unified “virtual” representation of the measured audience.
- Ensures cross-media alignment without exposing personal user data or imposing a universal ID scheme.
Private Reach & Frequency Estimator (PRFE)
- Builds on VPF outputs using secure multiparty computation (MPC), delivering cross-publisher or cross-channel reach/frequency aggregates while respecting user-level privacy.
- Allows for flexible local definitions of metrics; the voluntary Halo Framework does not define a single standardized currency or measurement standard.
Benefits & Limitations of the Halo Framework
Benefits
- Neutral: Allows for measurement of any media. No single media channel is inherently favoured.
- Privacy-Centric: Advanced cryptographic approaches support data minimization and anonymization principles.
- Flexible: JICs or measurement companies can choose to adopt or extend the components to match local contexts or enhance their existing measurement tools.
Limitations
- No Mandated Approach: Use is entirely voluntary and any implementation details (including metrics or final usage) are determined by local parties.
- Not a Currency: Halo does not define a uniform trading currency; it’s purely an optional open-source codebase.
The key to unlocking progress on cross-media measurement is to come up with a reliable means of combining traditional, sampled panel data with scaled campaign impression data. The Virtual ID model is a now-proven concept that helps us solve this problem.