Everything you need to know about HyphaMetrics

HyphaMetrics turns what people watch and do across screens into raw, real-time data. In consented U.S. households, our patented AI identifies content, ads, brands, platforms, and the person actually watching. Customers license the resulting dataset for their own models and products.

Modern behavior is fragmented across streaming, cable, social video, gaming, mobile devices, FAST channels, dynamic ads, shopping, and other screens. Much of the existing data stack sees only pieces of that behavior or relies on modeled and self-reported signals. Hypha observes the underlying behavior directly.

Hypha is the thread-like part of fungi that forms a mycelial network. The name reflects the idea of a connective data layer across a fragmented media ecosystem.

Rosters, smartphones, and smart watches tell us who is in the room. Off-the-shelf Android TV boxes capture signal, and Hypha’s models turn that signal into structured data about content, ads, brands, platforms, and viewing behavior. We do not ask panelists to keep diaries, press buttons, or sit in front of a camera.

Audio ACR identifies content by matching audio fingerprints. Hypha uses computer vision and other proprietary identification methods to analyze the screen signal itself, so the system is not dependent on audio alone. That allows Hypha to identify behavior and content that legacy audio-first systems can miss.

Hypha’s current data covers broadcast and cable, streaming services, FAST channels, social and video platforms, gaming, dynamic and programmatic ads, cross-device journeys, AI-generated content, and other observed screen behavior supported by the product.

Hypha combines the household roster with consented presence signals from smartphones and smart watches to determine who is in the room. The system does not use a camera pointed at the couch.

Every Hypha household is consented and incentivized. Hypha describes the licensed dataset as consented and anonymized.

Customers license structured Hypha data under enterprise contracts and can integrate it into their own models, analytics, products, and workflows.

Today, Hypha licenses one syndicated U.S. panel dataset to multiple customers under annual and multi-year contracts. Adding a customer does not require collecting a second copy of the underlying data. Next, Hypha plans to license its models and data rights through consumer-device deployments. Long term, population-scale data licenses become possible as the models move into media-delivery infrastructure.

Current and target buyers include major brands, media networks, ad agencies, adtech platforms, streaming platforms, agentic marketing platforms, strategy consulting firms, and quantitative hedge funds.

5,500 homes is Hypha’s current national benchmark-scale target and aligns with contracted customer delivery thresholds.

Legacy systems remain strongest in broadcast and cable. Hypha is built to observe modern behavior across streaming, social video, gaming, dynamic ads, FAST channels, cross-device journeys, and other environments alongside linear TV.

Comscore is a Hypha customer and build partner. It funded early model training and licenses Hypha data to calibrate its own audience models.

The moat is the combination of three issued U.S. patents, six years of proprietary person-level viewing observations, a consented operating panel, proprietary models, panel operations, and enterprise customers already using the data.

Hypha has three issued U.S. patents: No. 10,932,002; No. 11,570,513; and No. 12,238,372. On the public site, we describe the portfolio collectively as covering content identification, ad and brand detection, and viewer attribution.

Nielsen filed four patent matters against Hypha between November 2021 and April 2024. Two went to trial in July 2025 and resulted in jury verdicts of non-infringement. Nielsen voluntarily dismissed one matter before trial and dropped the remaining claim shortly before trial. The matters are resolved.