Camila Besseler, CMO, Analyzes How Synthetic Data Transforms Market Research — and Why Governance is Essential

Marketing was one of the first areas to feel the direct impact of generative artificial intelligence. Texts, images, videos, and scripts began to be created by anyone with access to a chatbot. But the use of AI in marketing goes far beyond what is visible on the surface.

There is a less visible — and far more transformative — universe that involves the use of synthetic data for market research, product development, and competitive intelligence.

What is synthetic data and why does it matter now?

Generating synthetic data is not pressing a button. It means designing, with AI support, statistical simulations that represent behaviors and attributes of a population. These data do not copy real samples; they emulate possible patterns and scenarios, preserving relevant correlations and variables.

The result? A statistical mirror of the real world capable of accelerating discoveries, reducing costs, and mitigating privacy risks.

As our CMO highlights, in an article published on Mundo do Marketing:

"The potential is real, but the euphoria can be dangerous. There is a race in marketing departments to show mastery over synthetic data, often without understanding its limitations."

What is behind this discussion, after all?

Professionals responsible for market intelligence face a dual pressure: doing more with less and increasingly protecting personal data. Robust research, which once justified generous budgets, has become too expensive and slow. In contrast, decision cycles have shortened and the demand for efficiency has grown.

It is in this context that synthetic data gains strength. They offer a scalable alternative for creating and testing hypotheses quickly and at low cost, without sacrificing analytical quality.

Market figures confirm the movement:

• Gartner (2025): Up to 60% of leaders may face critical synthetic data failures by 2027 due to lack of governance
• McKinsey (2025): Up to 75% of companies will use GenAI to generate synthetic data by 2026, compared to less than 5% in 2023
• Nvidia: Acquisition of startup Gretel demonstrates that the industrialization of synthetic data is underway.

In practice, synthetic data can be used for:

Research pre-tests — Validate questionnaires, wording, and logic before going into the field

Hypothesis validation — Test pricing, communication, and offer adherence with agility

Expand representativeness — Cover hard-to-reach niches, such as high-ticket corporate decision-makers or undersampled regions

Synthetic personas — Create "digital twins" that simulate behaviors in continuous testing of messages, service scripts, or sales scripts

With this, the traditional market research model begins to be challenged. What used to take months between collection, analysis, and validation can now be done in days.

Concrete Benefits vs. Risks

The benefits are clear: speed, cost reduction per iteration, coverage of rare niches, and native privacy. But, as Camila warns, "synthetic data is not a magic solution."

Main risks:

Model collapse — When models train on synthetic data repeatedly, degrading the quality of information. It is the equivalent of making copies of copies.

Overconfidence — Without statistical validation and comparison with real samples, conclusions may appear consistent but do not hold up in the field.

Re-identification — "Synthetic" does not mean "anonymous". Even with anonymization, there is a risk of re-identification of sensitive patterns.

Therefore, governance is the key word. The LGPD remains the regulatory umbrella, and the ANPD is already discussing specific guidelines for AI and data protection.

How to Implement Synthetic Data Responsibly?

Adopting synthetic data is not just a matter of generating simulations — it is necessary to have a well-defined strategy. The CMO emphasizes that the first step is always having clarity of purpose: knowing exactly what one wants to achieve with the use of data and how those results will be applied.

Next, it is essential to plan validation. No data, whether real or synthetic, can support decisions without a solid foundation. To ensure this, it is important that synthetic models are confronted with real data whenever possible, to validate the accuracy and reliability of conclusions.

The transparency is also fundamental. When using synthetic data in research or analysis, it is important to clearly communicate that part of the data is simulated, so that everyone involved in the process is aware of the limitations and implications of this approach.

Decisions: smarter, not just faster data

The future of market intelligence and marketing is intrinsically linked to innovation in how we collect, analyze, and use data. Synthetic data represents one of the great frontiers of this transformation, offering exciting possibilities but also demanding extreme care regarding governance and quality control.

As Camila highlighted, the competitive advantage in the future will not come from whoever adopts the technology first, but from those who know how to lead it responsibly and ethically. The key will be finding the balance between the speed provided by AI and the human intelligence that ensures the decisions made are, in fact, the best possible for the business and its customers.

Read the full article on Mundo do Marketing