How to use synthetic personas for customer research

Building user personas used to mean relying on market research and educated guesses. But with synthetic personas, there’s another way, one that’s less time consuming. These AI-powered profiles combine real user data with machine learning to create lifelike characters that act and think like actual customers. It's an approach that lets you test your ideas while keeping real user data private.

What are synthetic personas?

Synthetic personas are created by feeding vast amounts of real customer data into AI models. What you get is simulations rather than direct representations of real people, designed to mimic patterns and behaviors observed in the data.

The data comes from multiple sources: 

• Demographics
• Purchase records
• Online behavior patterns
• Social media activity
• Customer service interactions. 

The AI model analyzes these patterns to generate lifelike simulations that can respond to questions and scenarios much like actual customers would. 

Let's say you're designing a new fitness app feature. A synthetic persona, we’ll call her Sarah, created from data about millennial fitness enthusiasts can tell you she prefers workout tracking with social sharing options because that matches the behavioral patterns seen in the data. She might even point out specific concerns about privacy settings based on common user feedback in this demographic.

How are they different?

What makes synthetic personas different from traditional market research is their interactive nature. Rather than just describing customer segments, they can engage in simulated conversations. Or they can react to new product concepts and provide feedback. For example, a product team might present a new feature to a synthetic persona and receive specific responses about potential concerns or desired improvements.

Synthetic personas have important limitations however. While they're built on real data, they can't fully capture the nuanced emotional aspects of human decision-making. They also tend to provide overly positive feedback. This is a phenomenon known as AI sycophancy, where the AI's responses are excessively agreeable or flattering, often prioritising what it "thinks" you want to hear over providing a balanced perspective.

For this reason, they’re often better suited for early research and ideation rather than final validation of concepts.

Benefits of synthetic personas

Synthetic personas offer several advantages over traditional research methods, like …

Speed and scale

Teams can run hundreds of simulated interviews in minutes rather than spending weeks recruiting and interviewing real users. A study featuring over 1,000 participants found synthetic personas could match real participants' survey responses with 85% accuracy - nearly as reliable as how consistently people answer their own surveys after two weeks

Cost efficiency

Once set up, synthetic personas can be used repeatedly without ongoing recruitment costs or the need to schedule participant sessions and incentives.

Availability

Research can happen at any time without scheduling constraints or participant fatigue. Teams can run tests any time they need insights, whether it's late at night or during holiday periods.

Fine-tuning

You can adjust synthetic personas to match exactly who you're designing for. Whether it's tweaking their background, experience level, or specific behaviors, you've got all the controls to make your research scenarios hit the mark.

Range and reach

Synthetic personas give you access to perspectives from all kinds of users, from digital nomads in Southeast Asia to tech-shy seniors in rural areas. They're particularly valuable for understanding groups that would be difficult or time-consuming to recruit in traditional research.

Privacy compliance

Since they use anonymized data, synthetic personas help avoid data privacy concerns, as opposed to handling sensitive personal information from real research participants.

For example, when Apple was developing its iPad Pro campaign, early testing with synthetic personas might have flagged potential negative reactions to imagery showing creativity being "crushed." This kind of quick feedback could help teams catch issues before investing in expensive production and media buys.

These benefits do come with an important caveat however. Synthetic personas work best as a complement to, not a replacement for, research with real users. They're particularly valuable in early stages of development when teams need quick directional feedback or want to explore multiple concepts efficiently.

The downsides of synthetic testing

Testing with synthetic personas isn't perfect. Indeed, many scenarios can see it can steer you wrong. The most obvious problem is their constant positivity. Just ask one about your latest product idea and watch them generate a glowing review. Real people are rarely this agreeable.

There's also a concerning lack of emotional nuance. Let's say you're designing a new financial app. A synthetic persona can tell you people want to track their spending, but it won't capture that pit-in-your-stomach feeling when checking your balance before payday.

Everything hinges on your data too. Bad data built in means questionable insights out. If your training data only covers certain user groups or behaviors, your synthetic personas will have the same blind spots.

Not the all-in-one replacement for real people

Perhaps most worrying is how some teams are starting to treat synthetic personas as a complete replacement for real user research. There’s no doubt that the convenience is appealing. You remove recruitment hassles, get instant feedback, and have 24/7 access. 

But relying solely on AI-generated insights means missing out on those surprising, authentic human moments that often lead to breakthrough ideas.

A team using synthetic personas to test a new social feature may receive enthusiastic feedback about sharing workout achievements. But when they tested with real users, they might discover that people actually felt anxious about broadcasting their fitness journey. It’s this kind of emotional insight that just doesn't come through in synthetic testing.

Making synthetic personas work for you

Teams are finding ways to use synthetic personas by grounding them in high-quality human research data. The better your starting data, the more reliable your synthetic insights will be.

Early-stage exploration is where synthetic personas really shine. When you're trying to narrow down which ideas to pursue, you can easily test dozens of concepts. A gaming company using synthetic personas built from extensive player research could potentially evaluate 20 different feature ideas in a single afternoon, something that would take weeks with traditional testing.

They're also great for rehearsing research sessions. Before talking to real users, you can practice your interview questions with synthetic personas to spot potential gaps or awkward phrasing. Many researchers use this approach to sharpen their discussion guides, using patterns from previous successful studies.

Some teams use synthetic personas as an always-on research buddy, but only after establishing a solid foundation of real user data. When a random question comes up in a design meeting, they can get a perspective based on actual user patterns rather than making assumptions. It keeps genuine user needs in the conversation, even between formal research rounds.

A new twist for today’s research

Synthetic personas represent an exciting new era in research. While they can't replace real user insights outright, they offer valuable benefits for exploration and idea validation. Used thoughtfully alongside traditional research methods, synthetic personas can help teams work faster and smarter while keeping user needs at the center of their process.