×

The Trade Desk's Paula Bacariza Perez on Identity Graphs & The Graph of Graphs

←Back to Index

by Dana Lee on 19th Feb 2026 in TRADERTALK

In this Trader Talk TV episode, Paula Bacariza Perez, general manager of data & strategy in EMEA at The Trade Desk, joins ExchangeWire COO Lindsay Rowntree at the legendary whiteboard to discuss identity, identity graphs, and the graph of graphs.

Paula also explains the benefits of probabilistic data and why true omnichannel marketing requires a robust identity solution. By embracing an interoperable, multi-identifier strategy and trusting technology partners to manage the underlying complexity, advertisers can overcome the limitations of single-channel views and unlock the full potential of their campaigns.

Omnichannel, Multichannel, & Identity Graphs

A multichannel campaign uses various channels like TV, mobile, audio, and desktop; each use separate identifiers (CTV IDs, MAIDs, cookies). An omnichannel campaign uses a common identifier to follow the user across the channels, providing a unified view of the consumer journey.

An identity graph operates on three main constructs: devices, linked to device IDs; personas, a group of devices belonging to a user; and households, linking devices and personas based on shared patterns like IP addresses.

Probabilistic Vs. Deterministic Data

Devices are a deterministic construct, while personas and households are probabilistic, based on assumptions and deterministic links. However, the tendency to ignore probabilistic data can often create challenges in building a true omnichannel strategy.

Deterministic data may be technically infallible, but it does not allow for frequency capping, leading to wasted spend on unnecessary impressions. While not 100% accurate, probabilistic data helps reduce such waste by providing a better view and enabling scalability.

The "Graph of Graphs"

A graph of graphs connects multiple individual graphs through identity resolution, linking devices, personas, and households to all the different identifiers (eg. ID5, RampID, emetriq). This enables interoperability, solving the shortcomings of single identifiers by leveraging the best of many.

A key feature is translating identifiers back for reporting. For example, if an advertiser provides a RampID but the campaign converts on an impression with an ID5 identifier, the graph will report the conversion back using the original RampID. This ensures the advertiser can recognise personas and measure performance accurately within their own system.