An audience graph is a database that links the people and households a media company or platform can reach to what is known about them: which devices they use, what they watch or read, and which demographic or behavioural traits apply. It goes one step beyond an identity graph, which only records which identifiers belong together. An audience graph attaches attributes to those clusters so they can be grouped into segments, sold to advertisers, matched against an advertiser's own customer data and measured after a campaign. It exists because large media owners, above all television and streaming groups, hold first-party data scattered across dozens of services, and an advertiser can buy against that data only once it has been joined into one structure.

The term is a label rather than a standard. No industry body defines it, and companies apply it to different things. The dominant use, and the one this article follows, is the first-party graph a media owner builds from its own properties, of which the Walt Disney Company's Disney Audience Graph is the most prominent example.

How the graph is built

In data terms, a graph stores nodes and the edges between them. In an audience graph the nodes are typically households, individual profiles, connected TV (CTV) devices, mobile advertising IDs, browser cookies, hashed email addresses and IP addresses. Edges record that a device sits in a household, that a login belongs to a profile, or that a household watched a particular programme. Attributes such as age band, presence of children, genre preferences or purchase categories hang off the nodes.

The inputs come from the media owner's own operations. A streaming service contributes subscriber accounts, logins and viewing logs. A linear broadcaster contributes set-top box or automatic content recognition (ACR) data, which identifies what is on screen in televisions that report it. Ad servers add exposure logs. Third-party demographics are often layered on top. Disney says its graph is "built, not bought" and draws on linear, digital and streaming consumption, and lists two scale figures on its advertising website as of 2026: 115 million households and 450 million unique audience identifiers.

From the graph, data science teams derive segments. Disney Select, the company's segment product, offered more than 1,800 segments built from "more than 100,000 audience attributes" according to a Disney fact sheet dated March 1, 2022. A January 2023 Disney release put the count at nearly 2,000.

Where it sits in a campaign

An audience graph touches four stages of a buy.

Planning. Advertisers analyse the graph, usually inside a clean room, a controlled environment where two parties' data can be combined for agreed queries while raw records stay hidden. Disney paired its graph with a clean room built on Snowflake, according to VentureBeat in February 2022.

Matching. An advertiser uploads its own customer list, which is matched against the graph to find overlapping households. Because the brand and the media owner rarely share an identifier, a bridge is needed. Disney announced integrations with LiveRamp and Experian in February 2023, alongside The Trade Desk's Unified ID 2.0, according to AdExchanger. In June 2024 it introduced Bridge ID, which links its graph to third-party identity systems. The share of records recognised, the match rate, is the headline metric at this step.

Activation. Matched or pre-built segments are attached to inventory. Direct and programmatic guaranteed deals were the original route; since November 1, 2023, Disney+ segments from the graph have also been available through private marketplaces and invite-only auctions across 30 demand-side platforms (DSPs), according to MediaPost.

Measurement. Exposure logs are joined back to the graph, and to partner data, to count reach, frequency and outcomes. Disney integrated VideoAmp's TV viewership data into its clean room in December 2022, according to Marketing Dive.

Origin and evolution

The vocabulary comes from graph theory in mathematics, but its marketing use began with Facebook. At the company's F8 developer event on May 24, 2007, Mark Zuckerberg described the "social graph" as the network of real connections between people, and said "the social graph is at the center of that" shift in how information spreads, according to Facebook's announcement. Ad technology then borrowed the word for the cross-device device graphs of the early 2010s and the broader identity graphs that followed.

Exchanges were among the first to sell audiences built this way. In May 2019 OpenX described OpenAudience, which moved identity matching into the exchange using Tapad and LiveRamp data and claimed reach of 240 million unique people a month. "Everything that we do resolves to a person," Todd Parsons, then chief product officer, told AdExchanger.

Media owners then built their own. Disney's data work began roughly five or six years before 2022, according to VentureBeat. Disney Select launched around early 2021, and its clean room around late 2021. NBCUniversal revealed NBCU ID in March 2021 and in January 2022 described 150 million person-level and 50 million household identifiers, according to Marketing Dive. Disney extended its graph to Latin America in June 2024, the first region outside the United States, and in January 2025 launched Disney Compass, a data platform that sits over the graph and clean room.

Why it matters to marketers

Third-party cookies never covered television, and a growing share of video advertising now runs on logged-in streaming services. For buyers, a media owner's audience graph is often the only way to target a defined group across that owner's linear and streaming screens, and to count unduplicated reach across them. Disney Advertising reported a 56% rise in campaigns using audience data and 110% revenue growth in that business, according to VentureBeat in 2022.

PPC Land coverage shows how central the graph has become to Disney's sales story. When Disney opened ad-supported Disney+ in Latin America, the expansion brought its Audience Graph, clean room and Bridge ID outside the US for the first time. The company's streaming advertising results in early 2026 were framed around Disney Compass integrating the graph and clean room into one interface, and Compass became a route for Disney inventory bought through Amazon DSP.

Limitations and disputes

Size claims are self-reported and inconsistent. Disney's own figures have moved with the context: 200 million people reached in 2022, 235 million unique viewers and 110 million households in October 2023 (per MediaPost), and 115 million households and 450 million identifiers on its website in 2026. Each counts a different unit, and none has been publicly audited.

Match rates are vendor claims. Disney said in 2023 that clean room match rates were at least three times higher than traditional onboarding. AdExchanger noted the figure came from Disney and was not independently verified.

Graphs do not travel. Each media owner's graph and clean room is built for its own inventory. AdExchanger observed in 2023 that results from Disney's clean room could not be compared directly with other publishers'. Buyers working across Disney, NBCUniversal and others face several incompatible graphs, which reproduces the walled-garden problem at a smaller scale.

Household links are often weak. Many graphs rely on IP addresses to tie CTV devices to homes. A Truthset analysis for the Coalition for Innovative Media Measurement found IP-to-postal linkages accurate only 13% of the time, and a 2026 Adstra study found standard IP targeting missed its household roughly three times in four. First-party login data mitigates this, but only for signed-in viewing.

Privacy is configured, not automatic. Audience graphs are usually sold through clean rooms, which the US Federal Trade Commission (FTC) said on November 13, 2024 are "by default" not privacy-preserving. PPC Land reported the warning as a signal that clean rooms are no privacy silver bullet.

The term is loose. Some vendor glossaries, such as Blasto's, define an audience graph as a structure that "connects multiple identifiers" into a single view of a user or household, which is the definition of an identity graph. Disney and most streaming groups use it to mean identity plus attributes. Which sense applies depends on the vendor using it.

Not the same as

Identity graph. An identity graph answers which identifiers belong to the same person or household. An audience graph adds what those people watch, buy and are like, so it can produce segments. Many audience graphs contain an identity graph.

Device graph. A device graph links devices and browsers, often probabilistically. It is a narrower component of both.

Social graph and taste graph. Facebook's social graph maps connections between people, and Pinterest's Taste Graph maps users to interests. Both are platform-internal and drive recommendations and targeting inside one service rather than being matched with outside data.

Audience segment. A segment is the output, such as "sports fans with children". The graph is the structure from which segments are cut.

Recent developments

Disney's fiscal second-quarter 2026 report described engagement signals from Verts, its vertical video feed, flowing into Disney Compass, which widens the data the graph can draw on. In September 2026 Disney pitched its Audience Graph and Disney Select to Australian buyers at its upfront in Melbourne, months after Disney+ Standard with Ads launched there, according to Mediaweek.

Ownership of the identity plumbing beneath many graphs is also shifting. Publicis agreed to buy LiveRamp for $2.5 billion on May 17, 2026, noting that Disney, NBCUniversal and Netflix have integrated LiveRamp's RampID into addressable inventory. LiveRamp shareholders approved the merger on August 17, 2026 while rejecting the executives' pay package. As of October 2026 the deal had not been reported as closed. Platforms outside television are building household-level graphs too: Pinterest's purchase of tvScientific brought household graphs used to link CTV exposure to later purchases.

Timeline

  • May 24, 2007: Mark Zuckerberg introduces the "social graph" at Facebook's F8 event
  • May 3, 2019: OpenX describes OpenAudience, an exchange-level identity and audience product claiming 240 million people a month
  • Around early 2021: Disney Select launches
  • March 2021: NBCUniversal reveals NBCU ID
  • Around late 2021: Disney's clean room launches
  • January 5, 2022: NBCUniversal details NBCUnified with 150 million person-level and 50 million household identifiers
  • February 18, 2022: VentureBeat profiles the Disney Audience Graph and its Snowflake clean room
  • March 1, 2022: Disney Select reports more than 1,800 segments from over 100,000 attributes
  • December 5, 2022: Disney integrates VideoAmp TV viewership data into its clean room
  • January 25, 2023: Disney reports nearly 75 advertisers using its clean room
  • February 7, 2023: Disney adds LiveRamp and Experian integrations alongside UID2
  • November 1, 2023: Disney+ opens audience graph targeting through private marketplaces across 30 DSPs
  • June 19, 2024: Disney introduces Bridge ID
  • June 24, 2024: Disney's Audience Graph and clean room launch in Latin America
  • November 13, 2024: FTC staff say data clean rooms are not privacy-preserving by default
  • January 6, 2025: Disney launches Disney Compass
  • June 17, 2025: Disney inventory announced for Amazon DSP
  • May 17, 2026: Publicis agrees to acquire LiveRamp for $2.5 billion
  • August 17, 2026: LiveRamp shareholders approve the Publicis merger
  • September 10, 2026: Disney presents its Audience Graph at its Australian upfront

Summary

Who. Audience graphs are built mainly by media owners, led by Disney, whose Disney Audience Graph underpins Disney Select, its clean room and Disney Compass, and by NBCUniversal and other streaming groups. Identity providers such as LiveRamp and Experian supply the bridges that let advertisers match their data to them.

What. An audience graph links households, profiles and devices to attributes such as viewing, demographics and purchase behaviour, so that segments can be built, matched with advertiser data, activated and measured. It contains, but goes beyond, an identity graph.

When. The "graph" vocabulary entered marketing with Facebook's social graph in May 2007. Exchange-level audience products appeared around 2019, and media owner graphs took shape between 2021 and 2023, with Disney expanding internationally from June 2024.

Where. Graphs sit inside media owners' data platforms and are accessed through clean rooms, then activated through direct deals, private marketplaces and DSPs across linear, streaming and digital inventory.

Why. Cookies never reached television, and streaming viewing is logged in. A media owner's graph is often the only way to target and count audiences across its screens, though its size and match claims are self-reported, its links can be weak, and no two graphs interoperate.