Xandr

Xandr Builds an Identity Graph for Millions of Customers with Entity Resolution using TigerGraph

Xandr is the advertising and analytics division of AT&T’s WarnerMedia. The company operates an online platform called Community that connects advertisers, publishers, and consumer media brands. Community is built on Xandr’s powerful technology and consumer insights, enabling buyers and sellers to drive better business outcomes. Xandr was founded in 2018, employs approximately 1,800 employees, and is headquartered in New York City.

Graph + AI Sessions

The Challenge

Xandr wanted to create a digital advertising platform, Community, that would include over 15 WarnerMedia properties such as Cinemax, CNN, HBO, TNT, and more. However, since each of these brands has millions of consumers, disambiguating data in the proposed new platform would be a challenge. Not only that, Xandr wanted to combine all of this information in a way that would enable a better platform for advertisers, one capable of creating personalized commercials, with different promotions being delivered to individual viewers simultaneously. This would be an industry-first

Xandr

The Solution

TigerGraph is enabling Xandr to merge data from silos across the WarnerMedia universe and build the first, and largest, identity graph of its kind in the advertising industry. TigerGraph’s ability to integrate datasets and resolve ambiguous entities at scale is allowing Xandr to uncover insights hidden in its data. Specifically, Xandr is able to identify attributes associated with people, devices, households that provide marketers the ability to target audiences with customized commercials aligned to their interests and needs, resulting in better advertising performance.

The Results

Xandr’s SSE2 Team Lead (Associate Director), Data Science Engineering, Chinmay Nerurkar, says that identity graphs are important to advertisers as they stitch together different identifiers into a unified view of people, households and devices to enable cross-device and converged addressable advertising. To this end, Xandr has developed an identity graph with over five billion vertices and seven billion edges implemented with a ten node distributed cluster, each node with 400 GB RAM and 48 cores using TigerGraph. This is enabling Xandr to increase revenues by attracting companies to advertise on its platform, instead of using competitive products or other outmoded approaches.

ENTITY RESOLUTION USING CENTRALITY ALGORITHMS. CENTRALITY IS A MEASURE OF A VERTEX’S IMPORTANCE.

Source: “Identity Resolution at Scale” by Chinmay Narurkar and colleagues, presented at Graph + AI World 2020 conference. Watch the session: https://info.tigergraph.com/ graph-ai-world-xandr