The synthetic data conversation has been filled with hype, bold claims, and skepticism. But what happens when those claims are put to the test?

In our latest Between Two Joels session, AI experts Joel Anderson and Joel Armstrong unpacked their new whitepaper, When Synthetic Data Works, offering an independent validation of synthetic data boosting. Rather than debating theory, they shared hard evidence, showing how Fairgen’s method performed on real-world datasets, what it means for smaller subgroups, and where its limits lie.

Here’s what we learned together:

• The difference between fully synthetic data and augmented synthetic data.

• Transparent, statistically rigorous results on when boosting improves accuracy and effective sample size.

• A framework to know when synthetic augmentation is worth using (and when it isn’t).

• Practical implications for insights leaders navigating small-sample challenges.

Unlike past sessions that introduced synthetic data, this one validated it—challenging assumptions and clarifying when and where boosting really adds value.