Significantly reduce your sample requirements when collecting data by applying predictive augmentation
Smart
Use reliable predictive intelligence that is trained and monitored on your data
Economic
Reduce sample sizes needed to run your ongoing consumer trackers and surveys
Flexible
Works with any workflow, whether in-house, with agency or digital panels
Powerful
Analyze segments once considered too hard or slow-to-reach
Technical validation
Boosting consumer research about smart speakers
Fairgen's synthetic niche dataset performed better than 2x across segments, questions, and answers — and it was trained on only 44% of the ground truth.
Could Fairgen reliably augment all underrepresented segments?
Yes, Fairgen could reliably augment all 21 niche segments (+2x) using predictive synthetic data.
We validated with a large consumer expectations survey dataset of 6,034 rows. Fairgen was trained on a fraction of it (1,000 rows) and the rest served as ground truth.
The augmented dataset presented a very low MAE, right below 3%, and the quartiles were close together, showing that Fairgen could remove edge cases.
It performed better than a real sample twice as big as the trained one, showing that Fairgen could effectively predict and augment each niche by at least 2x.
What results did Fairgen provide when zooming in on one sub-segment of interest?
When looking only at one sub-segment of interest, in this case adults between the ages of 20 and 30, all questions maintained low distances to the ground truth (with a median below 5%), much better than 2x.
