Are synthetic audiences the future of marketing testing? 

Focus groups can be an important marketing tool, providing real-world insights straight from customers’ mouths. However, they also have limitations in speed, cost and scope. Fortunately, a solution may be on its way: synthetic audience testing.

Synthetic audience testing involves creating digital twins or avatars of customers that can answer marketers’ questions. 

“Synthetic audience testing would allow me to create that audience using real data,” said Camille Manso, a partner at innovation advisory firm Silicon Foundry.  

Unlike synthetic data, which is artificially generated, synthetic audiences are constructed from real-world data such as demographics, purchasing behaviors, etc. This is rendered in the form of digital avatars representing audience segments that marketers can question about preferences.

“You could ask them the five w’s,” she said. “Why might you be looking for a new supplement? When would you purchase it? When would you incorporate this into your daily routine? How might you envision the product?”

A ‘person’ contains multitudes

Each “person” in the audience represents a different demographic group and not an individual. So, they could be further sorted by geography, income or whatever else you have data about. 

“For creating those profiles, a lot of that is still going to be the human defining what the segment consists of,” said Manso. 

Dig deeper: Finding the sweet spot between relevance and discovery in B2B lead nurturing

As with everything in marketing, much will depend on the amount and quality of data used to construct the audience. 

“You need to be prepared today to take advantage of the opportunities that are going to be available in the next one to three years,” she said. “So, what data do you have available to you? How organized is your data? Because these models and AI tools are only going to be as good as your data going into them.”

In the development stage

Right now, synthetic audience testing is an “over-the-horizon” technology, albeit one with huge potential. It could replace focus groups and surveys or supplement them. These methods aren’t enough by themselves. Focus groups are time-consuming and costly, and a small sample size can distort results. Consumers are less interested in participating in polls and surveys and aren’t always truthful when they do.

By comparison, synthetic audience testing could reduce time to insights. Instead of weeks spent recruiting, conducting and analyzing focus groups, marketers would be able to pose questions to their synthetic audience and receive responses immediately. 

“You can’t iterate quickly if you’re running a focus group,” Manso said. “Whereas if you’re talking to an AI, and you’re behind a computer, you’re changing things whenever you want. Like, ‘What if we put it in white packaging? What if we put it in sustainable packaging? What would your reaction be?'” 

Easy to test

Manso suggested that companies can quickly test a synthetic audience’s usefulness by running the same questions with it and traditional focus groups.

Currently, the focus is on grouping individuals based on large language models trained on specific user data rather than achieving true individual-level understanding or a “segment of one.”

Looking ahead, a key goal of synthetic audience testing is nailing down microsegmentation, which aligns with the broader trend of hyper-personalized marketing. Interestingly, synthetic audiences could be used to understand when personalized marketing gets too hyper by gauging what consumers might perceive as “creepy.”

Dig deeper: Is A/B testing dead?