Media Buying Briefing: Synthetic audience adoption by agencies and clients takes off

According to Paul Frampton-Calero, CEO of Goodway Group, the solution will allow Goodway staff to run incrementality tests in a simulated environment, before verifying results with small-scale media activations. The hope is to execute against more accurate media plans, armed with rapidly generated test data.

Synthetic data — also called digital twinning or “audience cloning” — refers to the use of generative AI to create audience profiles, consumer cohorts or virtual focus groups from large sets of survey data. While approaches to input data, and the precise marketing applications the tech is being yoked to, are many, techniques based around the use of synthetic data are catching on fast.

Frampton-Calero said the solution could patch “gaps” in traditional research methodology with predicted responses taken from a synthetic audience — a little like repairing notches in a wall with filler. “If you’re trying to understand a particular product line or a particular region, the data is usually quite limited,” he noted.

Having access to a dataset representing a given population, he said, allowed the company to model their responses to marketing campaigns or activations. Furthermore, it gave staffers a way to test incremental increases of media spend without leaving a given market or region inactive to provide a control.

“We can query the data and say: if we spend at different levels, and push awareness or share of voice up, what would happen? Against a synthetic audience, we can simulate it without actually having to run a test,” Frampton-Calero explained. 

Goodway had been working with Arima since August, he said; the company also works with Optimal to create synthetic audience personas.

Arima’s data is based upon datasets such as censuses, figures from the U.S. Labor Bureau, Nielsen survey data, or bespoke consumer surveys; the synthetic “populations” are refreshed with new data at regular intervals. Goodway will also draw from first party data held by clients.

Goodway is not alone in pushing ahead with synthetic audience adoption. Ice cream-maker Magnum, shoe brand Vans and pharmaceutical advertiser Kenvue, for example, have been working with Vurvey Labs to use synthetic personas based on both public data and qualitative video surveys with real consumers, to inform product development and strategic insights. 

Chad Reynolds, CEO of Vurvey Labs, said the qualitative inputs meant clients got richer insights from its synthetic research. 

“LLMs give you the top of the bell curve, the most common answer,” said Reynolds. “We focus on the things that aren’t in language models… We’re asking real people around all different facets of their life.”

Joy Doyin Ogunneye, global innovation and brand communication lead for Kenvue skincare brand Aveeno Body, said use of synthetic research had sped up work on product and strategy development. Previously, she said, a strategy brief might have taken 30 days to write, but that time had been shortened to three days, she said.

“Good things take time. But what Vurvey has done is expedited that process,” she said. Doyin Ogunneye said that development of Aveeno’s Sensitive Skin Scents, launched earlier this year, was aided by Vurvey’s synthetic findings.

Among the Cincinnati, Ohio, company’s backers is Havas, which was part of a $8.5 million funding round concluded last November; the agency group is a “preferred partner,” according to CMO Heather Bollinger.

Meanwhile, beauty brand Sacheu has been working with Limitless Labs to create AI audience personas for use in virtual focus-group testing, this time based on the posts and online activity of its creator roster. “We ingest and analyze thousands of TikTok and Instagram videos,” explained Dale Janee, vp of business development at Limitless Labs.

While the number of agencies and clients working with the likes of Arima, Limitless and Vurvey’s applications is picking up, these applications aren’t new to the industry. 

Earlier this year, WPP began using a synthetic data solution based on charity donations data to directly inform media buys in the U.S. And Dentsu has been working with startup Evidenza to incorporate synthetic data into its media planning processes for over a year. Firms like Brandtech’s Jellyfish, and Stagwell agency Code & Theory, for example, have been creating audience personas based on synthetic data and AI agents for several years.

Stef Hoffman, head of enterprise transformation strategy at Code & Theory, said teams commonly used personas to guide the development of creative assets from pitch, to brief on through to production. “It makes it so that your customer can be omnipresent,” she said. 

Code & Theory’s approach, which was developed in partnership with Adobe, rooted audience personas in client-held first-party data first, and then added relevant third-party or public data sources as appropriate — rather than the other way around. She argued this granted the personas a better connection to the real customers buying a client’s products.

Julie Clark, svp of diversified markets, media and entertainment at TransUnion, suggested marketers take a similarly critical approach to synthetic personas or models based on non-proprietary data. Marketers should be wary of relying on personas built atop “garbage,” she said. “You have to start with the right foundation.”

Code & Theory’s years of using synthetic personas had also helped it learn the technique’s limitations, Hoffman noted. Despite their ability to speed up market research in media planning or strategy applications, users had to maintain a professional distrust of their findings. “It informs judgment, but it doesn’t replace judgment,” she said.

Despite those warnings, usage of synthetic audience tools is likely to proliferate. Frampton-Calero told Digiday his ambition was to use Arima’s tool to test solutions to marketing problems beyond media buying, including how and where brands sell their products, and how they price them.

“We’re trying to get it to not  just be about the media. Our ambition is to use it for answering marketing questions,” he said. 

Color by numbers

Fresh WARC figures confirm the ongoing migration of ad spend towards social video, as well as a moderation of ad spend growth next year. The researcher projected global ad spend to grow 11.9% to $1.34 trillion this year, following 10% growth in 2025 and 2024. That’s expected to slow to 8.4% next year, in part due to the lack of a major sporting event like the Olympics or World Cup, or the political spend associated with next month’s midterm elections in the U.S. Suzy Young, the head of WARC Media Data, said the spending reflected an “increasingly uneven economy” propped up by AI capital spending even as cost-of-living pressures increase. Some stats: 

• $394.6 billion: total social media ad spend this year, up 21.3% on 2025. WARC forecasts social ad spend will reach $500 billion in 2028.
• 20.7%: the proportion of ad spend coming from tech and electronics advertisers, the largest category in WARC’s forecast. Travel (19.3%) and autos (17.8%) were also large spenders.
• 59.7%: The proportion of global ad spend that will flow to the combined forces of Alphabet, Amazon and Meta this year, about $659.6 billion. It’s expected to rise to 61.5% by 2028.

Takeoff & landing

• Brandtech Group launched a data and AI practice called Brandtech Data. The new unit will sit above its Fifty-Five, DP6 and Jellyfish businesses and provide AI transformation, CX, measurement and foundational data services to clients. “In the AI era, you don’t need to own more data for the sake of it,” said Brandtech founder & CEO David Jones. “You need to be much better at putting the data you already have to work.”
• In acquisition news, private equity-backed marketing group SAMY acquired U.S. social agency Get Engaged; tems of the deal were not disclosed. And MSQ followed up on its recent absorptions of Big Spaceship and Arke with a deal to acquire media agency January Digital.
• Personnel moves: WPP hired Tyler Turnbull, global CEO of McCann Group, to be CEO of WPP Creative … RPA hired Renee Chez to be chief growth officer … Camphouse appointed Graham Brown, co-founder of Mediasense, as a non-executive director as it expands its AI capabilities.

Direct quote

“I was talking to one client, and they were saying, ‘If I just let this go and the agent is choosing to just hit this audience — let’s say it’s in streaming, and there’s some sort of live programming where it has a huge concentration. All of a sudden, will I blow through my budget in one day?’ Agencies are certainly grappling with what that looks like. We’re in a test and learn phase. It must have happened if people are worried about it. Things like what is the liability and the concern? And what are the protocols that you use?”

— Julie Clark, svp of diversified markets, media and entertainment at TransUnion, discussing agencies rolling out buying and planning agents with caution.

Speed reading

• Missed Advertising Week? Aren’t you’re lucky. But if you want to know the important stories that came out of it, check out Digiday’s coverage on Monday, Tuesday, Wednesday, Thursday and Friday of last week.  
• Speaking of agentic media buying and planning, Seb Joseph covered Monks’ efforts to let its agents run with minimal human oversight. 
• Michael Bürgi covered two news developments out of Omnicom: ressearch that dug into ad avoidance and what can be done about it, as well as an in-content advertising partnership with ad tech firm Rembrand.