Synthetic Buyer Persona: A Revolution in User Research or a Risky Bet?

The advancement of generative artificial intelligence has given rise to a new tool in user research: the synthetic buyer persona. This concept promises to accelerate the acquisition of insights, reduce costs, and improve scalability in strategic decision-making. However, is it truly a revolution or a risky shortcut that could compromise the quality of research? In this article, we explore the benefits and limitations of this methodology, evaluate its impact on the industry, and argue why a hybrid approach may be the key to balancing innovation and authenticity in user research.

A few weeks ago, I facilitated a session with my colleague Laura Polls at a Runroom LAB, the monthly event we organize at Runroom, on the possibilities that generative AI offers us in user research and specifically in the creation of the synthetic buyer persona.

It was a session where we presented our vision on the situation and the impact it could have on our industry.

The buyer persona, as practically everyone knows today, is a fundamental tool in marketing and product development. It represents an archetype of the ideal customer, based on research and data, which helps companies better understand their audience and make more informed decisions.

In the past year, we have witnessed the emergence of a new variant: the "synthetic buyer persona." Driven by advances in artificial intelligence, this concept promises to transform the way companies conduct user research. Deloitte, in its report “The Generative AI Dossier”, notes that one of the most interesting use cases for the consumer-focused industry will be providing next-level market intelligence through AI-conducted market research. This promises to generate synthetic data, simulate market scenarios, reveal insights, identify customer preferences, and create detailed profiles, such as Persona profiles.

The Synthetic Buyer Persona promises to transform user research through the generation of synthetic data, market scenario simulation, and insight identification.

The use of AI in user research is also rapidly growing, with 70% of researchers incorporating it into their workflows, according to the The 2024 AI in UX Research Report by User Interviews, reflecting a growing trend towards optimizing research processes.

Are we facing a true revolution or a shortcut that could lead us down the wrong path?

Next, I will try to explore the pros and cons of using AI as the core of research, focusing on synthetic buyer personas.

What is a Synthetic Buyer Persona?

A synthetic buyer persona could be defined as a language model (LLM) that has been trained to act like a real person. These language models have learned to act like people. Even researchers from Peking University (Jiang et al., 2023), claim that these develop a personality that guides their behavior, which can be induced by "prompting". Through AI, these models learn to simulate behaviors, preferences, and even personalities, allowing researchers to interact with them as if they were real users.

The data driving their development includes demographic and psychographic data, purchase history, online behavior, or social media activity, for example.

Synthetic data, meaning not created by humans, mimics real-world data. They are created through algorithms and simulations based on generative AI. A set of synthetic data has the same mathematical properties as the real data it is based on, meaning they are designed to replicate the statistical characteristics of an authentic data set, maintaining the same distribution, correlations, and patterns as the real data they are based on, but do not contain the same information, allowing them to be used in analysis without compromising privacy or relying on direct user information collection.

These language models can act as “virtual” subjects to whom specific preferences, information, or situations are assigned (for example, how much they value fairness, if they have a certain political ideology, etc.). Then, one can observe how they “respond” in various classic economic games and experiments. In different experiments, the models respond similarly to humans, although with nuances depending on the version of the LLM or the way the simulated personality is “configured,” according to the study by John J. Horton, Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus? (Cornell University, 2023).

While synthetic buyer personas offer speed and scalability, their lack of nuance and the risk of biases can compromise the authenticity of the insights.

Benefits of Using 100% Synthetic BP

• Speed and Flexibility: A 100% synthetic buyer persona can be created quickly, without relying on surveys or interviews with real users. This allows testing different scenarios, attributes, or market hypotheses in much less time. AI allows generating "synthetic data" and "simulating market scenarios," leading to a deep and quick understanding of the market, providing advanced market intelligence that allows obtaining deeper and faster insights through AI-conducted market research.
• Cost and Time Savings: By not having to recruit or coordinate study groups, the costs associated with traditional research (focus groups, interviews, etc.) are reduced. Additionally, the time to explore multiple profile variants is optimized. One of the biggest obstacles in user research remains participant recruitment, with 97% of researchers facing difficulties. The main issues include finding participants who meet the necessary criteria (70%) and the slowness in recruitment times (45%). Synthetic buyer personas can mitigate this problem.
• Continuous Iteration and Refinement: Being a "living" and adjustable profile, it is easy to update it with new information or expert opinions, allowing continuous refinement of marketing strategy or product development. This fosters data-driven decision-making and improves adaptation to market changes.
• Scale and Variety of Scenarios: One or several synthetic buyer personas with different characteristics (different motivations, locations, etc.) can be generated to cover various market segments. This facilitates the validation of new marketing or product strategies without additional investment in field research.

70% of researchers are already incorporating AI into their workflows, reflecting a growing trend towards optimizing research processes.

Cons and Limitations

Despite its advantages, the use of synthetic Buyer Personas also presents a series of drawbacks and limitations that must be considered:

• Lack of Nuance and Authentic Perspectives: Conducting direct research with real people provides nuances that AI or simulated data might overlook (e.g., specific fears, relevant anecdotes, authentic language, and unconventional expressions).
• Risk of Bias in Generation: Any AI model or method can incorporate biases present in the database or the algorithms used to create it. Interviewing real users allows contrasting data from multiple sources and reducing the risk of perpetuating prejudices or assumptions far from reality.
• Less Empirical Validation: With "on-the-ground" research, direct evidence is gathered on how the audience perceives, uses, or values a product or service. This provides much stronger validation of marketing or design hypotheses.
• Inability to Capture Human Spontaneity and Improvisation: In traditional qualitative research, open conversations, follow-up questions, and spontaneous interaction help uncover latent needs or hidden "pains" that had not been considered.

Seeing the benefits and cons of having a 100% synthetic BP, we are left with the feeling that it may work in some cases but not in many others. Therefore, we believe that the ideal solution lies in a hybrid model.

The key is not in choosing between synthetic or traditional buyer personas, but in adopting a hybrid approach that combines efficiency with depth and authenticity.

Hybrid Methodologies

Given this scenario, a promising solution is to adopt hybrid methodologies that combine the best of both worlds. This would break conventional boundaries between qualitative and quantitative research, giving way to a methodology that is neither purely qualitative nor purely quantitative.

Although models can be fed with data, it is also possible to explore the nuances and quality of experiences at a deeper and qualitative level.

These data could come from both qualitative and quantitative studies previously conducted with real people.

A hybrid approach allows leveraging the efficiency and scalability of synthetic buyer personas while maintaining the depth and authenticity of traditional research.

Therefore, we would be talking about creating a synthetic Buyer Persona, also relatively quickly, based on data previously collected through research with real people. This BP would be endowed with the same flexibility and speed and could respond with behavior supposedly validated with real people.

Synthetic Buyer Personas represent a promising innovation, but their exclusive use could lead us down the wrong path if not combined with real research.

In conclusion, 100% synthetic Buyer Personas represent a promising innovation in the field of user research. They offer undeniable benefits in terms of speed, cost, and scalability. However, it is crucial to recognize their limitations and the risks associated with their exclusive use. The lack of nuances, the risk of biases, and less empirical validation are aspects that cannot be overlooked.

In our opinion, the present and immediate future of user research probably lies in a hybrid approach, where synthetic Buyer Personas are used as a complement to traditional methodologies. By adopting this mixed approach, companies can leverage the power of AI to optimize their research processes without sacrificing the depth and authenticity that only human interaction can provide.

But we do not rule out that at some point generative AIs may completely replace real user research in certain contexts. We will be attentive to the next steps :))