WPP Tests Charity-Donation Synthetic Audiences for Buys

WPP Media, the media arm of the WPP holding company, is piloting a new approach to programmatic targeting: synthetic audiences built from charity-donation signals rather than individual-level identifiers. The experiment is a notable example of how agency holding companies are reworking their audience strategies as third-party cookies decline and privacy regulation tightens the constraints on person-level data.

For the video and programmatic audience at Adelerate, the experiment matters less for its philanthropic framing and more for what it signals about the mechanics of modeled, cohort-based targeting feeding into the open programmatic pipeline.

What "synthetic audiences" actually means

A synthetic audience is a modeled cohort — a statistically generated population that mirrors the behavioral characteristics of a real seed group without exposing or relying on the underlying individual records. In WPP's test, charity-donation behavior serves as the seed signal: patterns in giving are used to infer broader attributes (affinity, propensity, lifestyle indicators) that can then be projected onto a synthetic population and activated against programmatic inventory.

The appeal is structural. Rather than building audiences from deterministic identifiers (hashed emails, cookies, device IDs) that are increasingly fragmented or deprecated, a synthetic approach generates a probabilistic model that can be applied at the segment level. That makes it compatible with the direction the ecosystem is moving — toward cohort-based and modeled targeting that doesn't depend on tracking individuals across the open web.

Why this fits the post-cookie addressability problem

The timing is not accidental. With third-party cookie signal eroding and privacy frameworks (GDPR, CPRA, and tightening consent enforcement) raising the cost of person-level data, buyers need targeting approaches that degrade gracefully. Synthetic and modeled audiences are one of several techniques — alongside data clean rooms, seller-defined audiences (IAB Tech Lab's SDA), and contextual targeting — competing to fill the addressability gap.

For programmatic activation specifically, the question is how these modeled cohorts get expressed in the bidstream. IAB Tech Lab's Seller Defined Audiences and the broader audience taxonomy work provide a path for passing standardized segment signals through OpenRTB without leaking identifiers. A synthetic audience derived from donation behavior would ultimately need to map to some addressable taxonomy or PMP/deal structure to be actionable across SSPs and DSPs. That translation layer — from a modeled cohort to a biddable segment — is where the technical substance lives.

Implications for video and CTV inventory

Modeled cohorts are particularly relevant to CTV and video, where deterministic, cookie-based targeting was never the dominant mechanism to begin with. CTV environments already lean heavily on first-party data, content-based signals, and modeled household graphs. A synthetic-audience approach slots naturally into that world, where buyers are accustomed to activating against probabilistic and inferred segments rather than 1:1 identifiers.

If WPP's experiment scales, it could reinforce a broader pattern: agency-side audience modeling that produces privacy-safe cohorts on the buy side, then activated through standard programmatic plumbing (PMPs, deal IDs, and curated supply) into video and CTV inventory. That keeps the activation infrastructure unchanged while shifting the audience-construction logic upstream.

The open questions

Several practical concerns remain unresolved. Donation-derived signals raise sensitivity questions — charitable giving can correlate with protected or inferable attributes, so the modeling and any taxonomy mapping will need careful governance to avoid recreating the privacy risks synthetic audiences are meant to sidestep. There are also measurement challenges: validating that a synthetic cohort delivers genuine performance lift versus contextual or broad-reach baselines requires controlled testing, and modeled audiences are notoriously hard to verify against ground truth.

Scale and supply compatibility are open too. A synthetic audience is only useful if it can be activated efficiently against real inventory at meaningful reach, which depends on how cleanly the cohort maps into the standards (SDA, audience taxonomies) that SSPs and DSPs already support.

Why it matters

WPP Media's test is an early indicator of where large buyers are placing bets as deterministic targeting becomes harder. Whether or not charity-donation seeds prove durable, the underlying move — toward synthetic, modeled, privacy-conscious cohorts activated through standard programmatic pipes — is squarely aligned with the direction of the open ecosystem. For ad ops teams and platform leaders, the takeaway is to watch how these buy-side modeling experiments connect to bidstream standards, because that integration layer will determine whether modeled audiences become a practical replacement or remain a controlled pilot.

Stay on top of video ad serving and programmatic. Follow Adelerate.