Household ID

Conviva is the leading streaming data platform, bringing publishers the insights they need for effective quality of service, marketing, and advertising analysis. The platform is built atop the Conviva Sensor, which is deployed on publisher video players with more than 3.3 billion instances worldwide and collects comprehensive, real-time cross-device data providing reliable and accurate information about video and ad performance for your entire streaming audience.

Insights Straight from the Video Player

Conviva helps solve the identity and measurement challenges with the industry’s most accurate streaming metadata collected directly from the video player, including classifying viewers/devices into households and assigning household IDs. Household identities along with extensive metadata and demographics empower a variety of use cases, driving innovative approaches to advertising and marketing business goals.

Data Collection

Conviva collects the highest fidelity data from the Conviva sensor within the video player and sends the data to the Conviva backend for processing. Over 4 billion Conviva sensors have been deployed, collecting data on over 200 billion streams per year. Household IDs are created through the analysis of streaming behavior patterns aggregated across multiple apps and devices collected by the Conviva Sensor.

Metadata collected includes:

• Content

• Advertising

• Engagement

• Publisher audiences (segmentation)

• Device

• Location

• Demographics from 3rd data sources, such as TransUnion.

Household ID Built for Streaming Platforms

The Household ID links multiple devices to a single household based on anonymized publisher data that excludes cookies and mobile identifiers (mobile ad identifiers).

• Maps multiple application downloads to one household

• Generates Conviva Device ID (multiple apps-client ids identified as one device)

• Mapping applications at the device level and devices to households helps to future proof publisher’s businesses. Conviva’s proprietary device ID removes reliance on device manufacturers to pass on device-level data.

Building the Household ID

1. App level metadata collection: Conviva sensor collects high-fidelity metadata from each video session from the video players; session level data is assigned to a Client ID, unique for every app download.

2. Clean the data: Identify IPs that can be matched to a residential household.

3. Detect primary residential IP for each client ID: For each client ID, Conviva reviews a defined historical period and identifies the primary residential IP.

4. Cluster client IDs into households: Conviva’s proprietary machine learning graph algorithms cluster client IDs based on the identified primary IP. The output of the model is a mapping of household IDs with a group of client IDs.

5. Cluster client IDs into devices: Conviva’s proprietary machine learning graph algorithms cluster client IDs based on the identified primary IP. The output of the model is a mapping of device IDs with a group of client IDs, which can be associated to households.

Standardized Data Collection

Household ID makes it easier for publishers to unify viewer identity across all their streaming points of distribution, including brands, applications, and devices. Standard data collection across all end points enables building viewer and household profiles across all points of distribution. Household ID is also ready to link to the larger ad ecosystem to enrich other data sources with a valuable, viewer-level experiential data set.

Persistent

Household ID is built upon a multi-ID framework, and works in cases of rotating or missing IDs including:

• Client IP Address

• Device Information

• Time Stamps

• Client ID

Continuous model improvement

Every stream is monitored and fed into the household model for continuous improvement. Conviva can detect when someone within a household watches content outside of a primary location, when a viewer moves out of a household, or when a whole household moves.

Daily updated

• IP address changes (e.g., moving to a new home): The IP address-related feature in the predictive model is the ratio of shared IP addresses in the last 28 days. Even though the device IP address is changing, the IP shared device count will not change significantly.

• Viewer migrations (e.g., kids move out to college): The IP address and ZIP code level location-related features are the most important in the model. When viewers move, their location and IP address change and the model will reassign the viewer’s household.

Conclusion

The Stream ID supports publishers in addressing streaming identity needs for marketing and advertising use cases, such as improving content recommendation, audience targeting, streaming measurement, and gaining deep audience insights.