How Synthetic Sample in B2B research enhances data quality

Recently, there’s been significant buzz about synthetic research in B2B research—both excitement and concern. “I’m skeptical…is there any research showing that synthetic data is accurate?” This is what we’ve been hearing from the research community. 

At NewtonX we specialize in B2B research, and we’ve been putting synthetic sample through a rigorous testing program for months, committing both technology and strategy resources to find the right solution. Our findings are promising and unique: synthetic sample can be incredibly valuable when used appropriately, enhancing or augmenting, rather than replacing traditional methods.

What is Synthetic Research or Synthetic Sample

Synthetic sample is an innovative methodology generating quantitative research responses using advanced statistics and artificial intelligence. 

It’s not just a mirror of existing respondents or a simple reweighting technique (which wouldn’t reduce the margin of error). The AI model learns from real world patterns to create new independent responses with their own answer nuances and texture as if you would field completely new real life respondents, increasing robustness of the sample. 

With this breed of synthetic data, which we consider augmented, you can collect more data in less time, enabling faster GTM decision-making without shaking your confidence.

The Role of Synthetic Sample

Synthetic sample addresses some common challenges in B2B research, such as high cost and limited access to niche markets. While there is enough public information on all types of consumers to create synthetic respondents for generic B2C studies from scratch, B2B is a whole different beast. What makes B2B challenging is that it relies on knowledge that by definition isn’t publicly available and is very specific to each research study, so it’s essential to have a base to train the AI model. 

NewtonX research shows that synthetic data in B2B is most impactful when it augments real responses in a given survey. Hence, we found augmented data is a reliable subset of synthetic sample and the right approach for B2B research. 

To put it simply:

• Create synthetic respondents from scratch –> Mainly works for B2C because of all the publicly available data
• Create synthetic respondents based on existing real respondents data points in a custom survey –> Works for more complex B2B needs (Augmented data)

Quality Foundations Matter

Effectiveness of Augmented Data hinges on the quality of the underlying real-world data used to train these AI models. Using low-quality panels with fraudulent respondents to train AI models can lead to unreliable results—a classic example of “garbage in, garbage out.” High-quality data from verified professionals is crucial for training AI models to ensure that augmented samples are accurate and valuable.

Good news however–our A/B tests reveal that augmented data generated from high-quality, verified respondents matches the quality of a fully custom recruited sample with a 95%-99.5% statistical equivalence. This saves days or weeks of fielding time and reduces the CPI, allowing us to pass those savings to our clients. 

Insights from A/B Testing B2B Synthetic Research

Our tests compared three independent sample sources for the same audience specs:

1. Augmented sample trained on quality real-world data
2. Custom Recruited sample excluded from the training model (control group)
3. Panel Sample

Repeatedly we saw 95-99.5% equivalence between augmented data and completely independently sourced Custom Recruited Sample (control group) while Panel Sample showed a variance of 20-40%. 

This demonstrates that Augmented Sample, when properly generated, is a preferred alternative to Panel Sample as it can increase feasibility, save time and save money. 

Introducing NewtonX Augmented Data (NXAD)

Our latest innovation, NewtonX Augmented Data (NXAD), embodies our approach to integrating synthetic data with high-quality real-world samples. NXAD uses AI to generate synthetic respondents based on our verified data, offering several key benefits:

1. High-Quality Data Access: NXAD makes it possible for clients to access top-tier data even with tighter budgets, maintaining data integrity while reducing costs.
2. Avoiding Low-Quality Panels: Traditional methods often mix high-quality samples with less reliable panels to balance costs. NXAD provides a superior alternative by generating high-quality synthetic data, reducing reliance on subpar panels.
3. Meeting Niche Quotas: NXAD helps achieve specific, hard-to-reach quotas, enabling more detailed and granular analysis.

Embracing AI Thoughtfully

The integration of synthetic respondents into B2B research is transformative, and at NewtonX, we’re excited to be at the forefront of this evolution. While AI and synthetic data holds great potential, it must be used in conjunction with high-quality real-world data to be most effective.

As AI continues to shape the industry, a balanced approach will be crucial for leveraging its benefits effectively. Synthetic samples are a valuable tool for enhancing B2B research when used correctly. At NewtonX, we’ve always been an AI-driven company, applying automation and technology to B2B research, and this is a particularly exciting time. AI in B2B research is something to embrace, but thoughtfully.